Accepted Contributions

List of all Long papers accepted for RecSys 2026 (in alphabetical order).

• RESA Control Function Framework for Mitigating Position Bias in Learning to Rank Systems
  by Md Aminul Islam, Kathryn Vasilaky and Elena Zheleva

  Learning-to-rank (LTR) systems commonly depend on implicit feedback, such as user clicks, because it is easy to collect and can serve as a valuable signal of user preferences. However, directly optimizing ranking models using implicit feedback data often yields suboptimal performance because such data is inherently skewed by systematic biases. Among these biases, position bias is particularly pervasive: items ranked higher tend to receive disproportionately more interactions, regardless of their actual relevance. To address this, we introduce a novel two-stage framework based on control functions. In the first stage, we utilize exogenous variation from the residuals of the ranking process, which are then incorporated into a second stage click model to account for position-dependent distortions. In contrast to existing methods, our approach avoids explicit propensity estimation, supports nonlinear ranking models, and can be flexibly incorporated into any state-of-the-art ranking algorithm for position bias correction. We also propose a debiasing strategy for validation clicks that enables reliable hyperparameter tuning in the absence of unbiased validation data. Empirical results show that our method outperforms state-of-the-art position bias correction methods on both benchmark and real-world industrial datasets.

• RESA Redundancy Reduction Approach for Controllable Sequential Recommendations
  by Veronika Ivanova, Marina Munkhoeva, Ivan Razvorotnev and Evgeny Frolov

  Sequential recommendation must operate under long-tailed item distributions and popularity-driven concentration, often forcing practitioners to trade short-list accuracy against long-tail exposure. In this work, we study feature decorrelation as a mechanism for shaping representation geometry in dot-product sequential recommenders, and analyze how this, in turn, affects popularity-driven concentration. We propose a decorrelation-regularized training framework that augments next-item prediction with an auxiliary redundancy-reduction term, and instantiate it with BT-SR, which uses the Barlow Twins objective. To form label-consistent positive pairs without synthetic corruptions, we pair user histories that share the same next-item target. Beyond accuracy, we provide a geometric analysis showing how decorrelation suppresses shared low-rank directions in the user representation space that can give popular items a global scoring advantage, and we introduce a bucket-based alignment concentration metric to quantify this effect. Experiments on five public benchmarks show that BT-SR consistently improves next-item ranking quality, while the decorrelation strength acts as a simple control knob that reallocates accuracy across head and tail items, enabling accuracy–exposure trade-offs. Our analysis also reveals that the impact on head-vs-tail exposure differs across datasets, reflecting interactions between decorrelation and data temporal structure.

• RESAdaptive Retraining of Recommender Systems via Reinforcement Learning
  by Diego Russo, Valerio La Gatta, Claudio Spasiano and Vincenzo Moscato

  Modern recommender systems operate in dynamic environments where user preferences drift, new items arrive, and interaction patterns evolve, causing deployed models to become progressively stale. Retraining is essential to maintain recommendation quality, yet prior work has largely treated the how and when of retraining separately: adaptation strategies are evaluated under fixed schedules, while scheduling policies assume predefined updates. We formalize retraining as a sequential, resource-constrained decision problem that jointly determines when and how to update a recommender. Rather than introducing new retraining algorithms, our approach leverages existing strategies, including full retraining, fine-tuning, and sample-based updates, selecting the most effective action at each timestep. We introduce RTagent, an agent instantiated via reinforcement learning, which learns a meta-policy optimizing long-term cumulative performance under a global constraint on the number of retraining operations. Evaluation on MovieLens 1M and Yelp across three recommender architectures (SVD, CAFE, and NeuMF) shows that RTagent consistently outperforms static schedules, closely approaches full-retraining effectiveness while operating under the same budget constraint, and exhibits interpretable, architecture-specific retraining rhythms, demonstrating the benefits of sequential, strategy-aware retraining decisions.

• RESAddressing Cross-Stage Decoupling of Semantic and Collaborative Signals in Generative Recommendation
  by Jiayi Dan, Weijian Li, Yongqi Liu and Kaiqiao Zhan

  Generative recommendation reformulates sequential recommendation as autoregressive generation by encoding items into semantic tokens, enabling improved scaling capability and cross-domain generalization. However, existing generative recommender systems typically follow a two-stage pipeline, where item tokenization is largely dominated by textual semantics with limited incorporation of collaborative signals and interaction similarity, leading to code assignments that are misaligned with downstream generation. Conversely, the generation stage tends to overlook the original semantic information, as the code sequences are re-embedded based on interaction data. This cross-stage information decoupling limits semantic coherence and recommendation accuracy. To address this issue, we propose SCRec, a general framework that enhances cross-stage coherence through bidirectional information supplementation. Specifically, we introduce (i) collaborative-enhanced tokenization to explicitly inject textualized collaborative signals into semantic tokenization, without introducing additional alignment task, (ii) semantic-guided generation to dynamically recalibrate semantic priors with learnable code embeddings in generation stage, and (iii) manifold alignment to reconcile the geometric mismatch between the embedding space of discrete codebook indices and the dense continuous semantic space. These interrelated components form a general framework that aligns semantic and collaborative signals and enhances cross-stage information coherence, with minimal additional training and inference costs. Extensive experiments demonstrate the effectiveness, robustness, and generalizability of our proposed framework.

• RESAutomated Selection-based Mixture-of-Experts with Dual-stage Input–Target Pattern Learning for Sequential Recommendation
  by Xiaolin Lin, Weike Pan and Zhong Ming

  Sequential recommendation (SR) aims to predict the next items for users by learning the users’ representations from their historical sequences. In this process, most existing methods rely on a single globally shared encoder to model the relationship between input sequences and target items, implicitly learning input–target patterns. However, such implicit learning treats patterns uniformly and largely overlooks their intrinsic characteristics and differences among training samples. Through the empirical studies in this paper, we find that input–target patterns exhibit both conflicting and generalizable characteristics, which impose distinct modeling requirements. Neglecting these properties leads to suboptimal user representations and limited generalization. Motivated by these findings, we propose a novel MoE architecture, Automated Selection-based Mixture-of-Experts (ASMoE), with a dual-stage training scheme to address these issues. In our ASMoE, we introduce an automated expert selection mechanism to adaptively allocate selectable experts and accommodate diverse modeling requirements of input–target patterns. Furthermore, we develop a dual-stage training scheme to enhance our ASMoE for input–target pattern learning. The first stage performs initial learning over diverse patterns. In the second stage, we explicitly construct the potentially generalizable input–target patterns via a category-aware mask generator and a similarity-aware penalty, thereby facilitating the fine-tuning of our ASMoE towards generalizable knowledge. Extensive experiments on four public datasets demonstrate the effectiveness of our ASMoE. The source code and auxiliary material of our ASMoE are provided at https://anonymous.4open.science/r/ASMoE.

• RESBeyond Fixed Depths and Widths: Optimizing Textual Decoding Tries in LLM-based Generative Recommendation
  by Jingzhe Liu, Hanbing Wang, Jiliang Tang, Liam Collins, Tong Zhao, Neil Shah and Mingxuan Ju

  Generative recommendation (GR) is an increasingly popular paradigm in recommender systems, with a prominent line of work using LLMs as autoregressive backbones to predict the next item’s term IDs (e.g., titles or keywords). The success of autoregressive generation hinges on constrained beam search over a decoding trie to ensure that generated outputs correspond to valid items. However, current research predominantly focuses on generating more comprehensive term IDs to describe items, while largely neglecting the structural design of the decoding trie formed by these terms. This can lead to a trie that is poorly suited to beam search, which degrades performance. To address this, we examine the effectiveness of term IDs from the perspective of decoding trie optimization. Through empirical and theoretical analyses, we identify two desirable properties for a highly performant trie: (1) adaptive and variable ID length, enabling items with varying semantic richness to be represented by IDs of appropriate lengths, and (2) constrained branching factors, especially at shallow levels, which drastically improves the success rate of constrained beam search. Motivated by these properties, we introduce BONSAI: Branching-Optimized Node Structure for Adaptive Identifiers, a novel framework that co-designs textual term IDs and their underlying decoding trie. BONSAI extracts recommendation-informative words from item metadata and employs a minimum set cover formulation to recursively build a trie that satisfies the above properties. Experiments reveal that BONSAI achieves up to a 21.6% relative improvement over state-of-the-art baselines. Further analyses confirm the crucial role of our proposed properties, and demonstrate their generalizability to be applied to enhance the performance of other term ID methods.

• RESBiLPR: Bidirectional Teacher-Student Agent Interaction for Context-Aware Learning Path Recommendation
  by Zejun Chen, Weiwei Chen, Suojuan Zhang, Zhi Zheng, Dawei Jin, Ziwei Zhao, Tong Xu, Jing Cui, Jiaqi Long and Enhong Chen

  Learning path recommendation is a critical component of intelligent education systems, aiming to plan a personalized sequence of learning resources for each student based on their cognitive state. Existing methods predominantly rely on unidirectional modeling for recommendations, failing to adequately capture the bidirectional interaction between teachers and students. This leads to a lack of feedback-driven adaptation and difficulty in forming an effective instructional closed loop. Furthermore, current learning path recommendations are often limited to static student-exercise matching. They cannot perceive and respond to dynamic learning contexts, which results in insufficient adaptability. This limitation stems from an inadequate consideration of key contextual factors, including real-time cognitive states, interaction history, exercise semantics, and knowledge structures. To address these issues, this paper proposes a Bidirectional Teacher-Student Agent Interaction for Context-Aware Learning Path Recommendation (BiLPR), which implements a bidirectional, dynamic, and synergistic process. Specifically, the Teacher Agent integrates domain knowledge graphs with semantic reasoning to thoroughly mine features of the learning context. This enables dynamic exercise adaptation and recommendation strategies underpinned by knowledge transfer. The Student Agent simulates the evolution of dynamic cognitive states and behaviors during authentic learning processes, providing feedback on its performance. This interaction establishes a novel iterative closed loop of recommendation, feedback, and reflection. Evaluated on two real-world educational datasets, Junyi and ASSIST2009, the proposed method significantly outperforms baseline models in recommendation effectiveness. The code is available at https://anonymous.4open.science/r/BiLPR-D16D.

• RESBilateral Intent-Enhanced Sequential Recommendation with Embedding Perturbation-Based Contrastive Learning
  by Shanfan Zhang, Yuan Rao, Yongyi Lin, Jia Lei, Linghan Zhang and Shuo Wang

  Modeling evolving user preferences from interaction sequences remains a core problem in sequential recommendation (SR). Recent work highlights intent learning for uncovering latent user motivations. Yet, existing methods either model intents within individual sequences or treat global intent signals as auxiliary supervision, limiting the explicit use of collective behavioral patterns and causing information isolation. Meanwhile, existing contrastive learning strategies are often costly and rely on suboptimal view construction, e.g., random sequence editing or weakly controlled model perturbations. We propose BIPCL, an end-to-end Bilateral Intent-enhanced, Embedding Perturbation-based Contrastive Learning framework. BIPCL integrates collective intent priors into both sequence- and item side representations via bilateral intent enhancement. Shared intent prototypes capture collective semantics from behaviorally similar entities and inject them into representations, alleviating information isolation and improving robustness. It further induces a non-separable cross-intent interaction, providing complementary sequence–item matching signals beyond unilateral intent modeling. To construct effective contrastive views, we introduce an embedding perturbation-based paradigm that directly perturbs structural item embeddings, yielding bounded and discriminative views while preserving temporal and structural dependencies. Compatibility studies across multiple CL-based SR backbones demonstrate the effectiveness of this paradigm beyond BIPCL. Extensive experiments show that BIPCL consistently outperforms state-of-the-art baselines. All code and datasets are publicly available at https://anonymous.4open.science/r/BIPCL-8E78/.

• RESBridge the Unseen Gap: Enhancing Non-overlapping Cross-domain CTR Prediction via Profile Retrieval
  by Jingyang Bin, Xing Tang, Wei Zeng, Jianan Su, Kaixin Shen, Jingtong Wu, Chaohua Yang, Kailiang Hao, Dugang Liu and Xiuqiang He

  Click-through rate (CTR) prediction is a fundamental task in industrial recommender systems. Cross-domain CTR prediction, which leverages data from a source domain to improve performance in a target domain, has emerged as a key strategy. However, most existing methods rely on overlapping users or items across domains to enable knowledge transfer, which fails in prevalent real-world scenarios where domains are functionally or geographically isolated (e.g., cross-country services). In this paper, we introduce a novel paradigm shift, from implicit representation alignment to explicit retrieval-based instance transfer. We propose LLM-PRIT, a framework for Large Language Model-generated Profile Retrieval & Instance Transfer. Our framework operates in three cohesive stages. First, it utilizes an LLM as a universal semantic interpreter to generate domain-agnostic, transferable profiles for users and items, encapsulating open-world knowledge. Second, instead of directly using these textual profiles, it employs them as semantic anchors to retrieve the most relevant historical instances from the source domain. This step explicitly establishes cross-domain correlations while avoiding the modality gap. Finally, it transfers knowledge by efficiently fine-tuning the target CTR model on the retrieved instances, preserving the model’s inherent feature-interaction capabilities. We conduct extensive experiments on a public and a real-world industrial dataset. Both online and offline results demonstrate the effectiveness of our LLM-PRIT, bridging the unseen gap with open-world semantic information.

• RESCONGA: Continual Neural Gated Architecture for Long-History Sequential Recommendation
  by Hoang Vu Le, Tuong Bach Hy Nguyen and Bac Le

  Existing sequential recommendation models rely on absolute positional encodings and fixed context windows, producing two structural failure modes: out-of-distribution degradation on histories longer than the training window, and hard context limits that discard long-range interactions. We present CONGA (COntinual Neural Gated Architecture), which addresses these limitations through three contributions: (1) Rotary Positional Embeddings (RoPE) with a norm-preserving property (||R_m||_F = sqrt(d) for any sequence length), accelerated by custom CUDA kernels; (2) KromHC multi-stream fusion with exact doubly-stochastic mixing via Kronecker-product parametrization, where ablation confirms the expressivity gain arises from balanced gradient flow rather than additional parameters, together with a data-adaptive stream selection mechanism that prevents overfitting on sparse corpora; and (3) TITANS neural associative memory adapted to discrete-item recommendation—the first such proof-of-concept—via a two-phase training protocol with a structural forgetting-prevention property: the base encoder is frozen, preserving its short-sequence predictions, while Phase 2 only adds a learned memory term. Evaluated under a rigorous full-ranking protocol across four benchmarks (ML-1M, Beauty, Yelp, Steam), CONGA achieves state-of-the-art performance with gains up to +29.7% on ML-1M—the densest, longest-history dataset where all three failure modes are simultaneously active.

• RESCalibrating User Preferences for Cross-Domain Recommendation via Target-Guided Representation Mapping
  by Guohang Zeng, Jie Lu and Guangquan Zhang

  As an interdisciplinary field between transfer learning and recommender systems, cross-domain recommendation (CDR) leverages a data-rich source domain to overcome the data sparsity issue in the target domain. In this paper, we study the problem of user preference calibration in CDR when the source domain contains noisy interactions, an aspect overlooked in previous studies. We demonstrate that noisy interactions in the source domain introduce a calibration gap — a divergence between the user representations learned in the noisy source domain and the user’s true preferences — which leads to inaccurate preference transfer to the target domain. To address this, we propose a robust CDR framework called Denoising Cross-Domain Recommendation (DCDR), which incorporates a target-guided representation mapping mechanism. The intuition behind this component lies in leveraging the cleaner user representations in the target domain to construct an explicit mapping function, thereby deriving calibrated user representations that correct the distorted source-domain preferences toward their true, preference-aligned counterparts. Notably, the proposed DCDR method is agnostic to specific CDR models, making it a general framework applicable to various existing CDR approaches. Experimental results show that our method effectively calibrates user preferences and mitigates the impact of noisy preference transfer, outperforming existing single-domain denoising approaches across multiple real-world recommendation tasks.

• RESCoarse-to-Fine Long-term Interest Modeling for Generative Recommendation
  by Shiteng Cao, Junda She, Ji Liu, Bin Zeng, Chengcheng Guo, Kuo Cai, Qiang Luo, Ruiming Tang, Han Li, Kun Gai, Zhiheng Li and Cheng Yang

  Leveraging long-term user behavioral patterns is a key trajectory for enhancing the accuracy of modern recommender systems. Due to the quadratic complexity of attention mechanisms, existing GR models are typically confined to short interaction sequences. While pioneer works have attempted to adapt Search-based Interest Models (SIM) to the generative context, they typically overlook the inherent hierarchical distinction of SIDs. GR is fundamentally a coarse-to-fine generation task, where the initial SIDs (prefix) determine the broad semantic category and the subsequent SIDs (suffix) pinpoint the specific item. Thus, our core insight is that the prefix and suffix of SIDs require distinct long-term signal injections. To bridge this gap, we propose GLASS, a Generative recommendation framework that integrates Long-term user interests into the generative process via SIDTier and Semantic Search. For the generation of SID prefix, we introduce SID-Tier, a module that maps long-term interactions into a unified interest vector to enhance the prediction of the initial SID token. SID-Tier leverages the compact nature of the semantic codebook to incorporate cross features between the user’s long-term history and candidate semantic codes. Furthermore, for the generation of SID suffix, we present semantic hard search, which utilizes generated coarse-grained semantic ID as dynamic keys to extract relevant historical behaviors, which are then fused via an adaptive gated fusion module to recalibrate the trajectory of subsequent fine-grained tokens. Extensive experiments on two large-scale real-world datasets, TAOBAO-MM and KuaiRec, demonstrate that method outperforms state-of-the-art baselines. A two-week online A/B test on a short-video platform demonstrate that GLASS achieves significant gains in recommendation quality. Our codes are publicly available at this anonymous link to facilitate further research in generative recommendation.

• RESConsensus vs. Dissent: Dynamic LLM Modeling of Subjective Preferences in Group Recommenders
  by Cedric Waterschoot, Nava Tintarev and Francesco Barile

  Previous work in group recommender systems has demonstrated a sensitivity to the distribution of preferences within a group. Specifically, the selection of the preference aggregation strategy benefits from considering such group configurations. In this paper, we study whether LLMs are able to mimic this sensitivity and to select the ideal aggregation strategy and corresponding recommendation according to nuanced human perceptions of fairness, satisfaction, and consensus. We do this by fine-tuning Large Language Models (LLMs) on human survey data to serve as real-time judgmental models within the recommendation pipeline. Using a reasoning dataset distilled from DeepSeek-V3.1 and human ground truth assessments, we develop Judgmental Llama and Judgmental OLMo to simulate group assessments. Our pipeline successfully generates multiple recommendation candidates based on social choice-based aggregation strategies and dynamically selects the one that maximizes these predicted human-like evaluations. We further validate these suggestions in a user study (n=284) and find that our methodology achieved the highest scores for satisfaction and group consensus. Furthermore, we find that LLM judgments are most aligned with human perceptions of fairness, satisfaction and consensus when we also consider interaction effects between our LLM-based method and group configuration (e.g., minority or coalition). These findings give further support for dynamically adapting aggregation strategies to specific within-group preference distributions, and highlight the advantage of using LLMs for an adaptation that is aligned with subjective human judgments.

• RESDart: Adaptive Tweedie Likelihood Mitigates the σ Escape Route in Cross-Platform Popularity Prediction
  by Tomohiro Mimura

  Although predicting social media popularity is crucial for modern recommender systems, it remains a significant challenge due to heavy-tailed, zero-inflated, and non-negative target distributions. While architectural innovations have advanced the field, standard objectives such as Gaussian negative log-likelihood (NLL) and mean squared error (MSE) are often poorly matched to these distributions, creating a critical performance bottleneck. In this paper, we diagnose one concrete manifestation of this mismatch, which we refer to as the sigma escape route: under Gaussian NLL, the noise parameter sigma absorbs prediction errors on heavy-tailed data, preventing the model from improving its mean prediction mu. To address this issue, we propose replacing Gaussian NLL with an adaptive Tweedie likelihood, which ties variance directly to the mean and naturally handles non-negative, zero-inflated, and heavy-tailed targets in a single parametric family. Notably, our empirical results demonstrate that the Tweedie likelihood consistently improves the Spearman’s rho independent of the model architecture. We also introduce Dart, a purpose-built retrieval-augmented system that achieves state-of-the-art ranking across multiple platforms. Dart excels in cold-start scenarios, substantially outperforming meta-learning baselines without requiring target-platform data. Furthermore, our downstream evaluations confirm that these ranking improvements translate directly into practical recommender-system tasks.

• RESDP-Rec: Towards Dynamic Patching for Efficient Long-Sequence Recommendation
  by Dwipam Katariya, Thomas Caputo, Akshat Shreemali, Juan Manuel Origgi, Pranab Mohanty, Nam Nguyen, Kalanand Mishra, Nikita Seleznev and James Montgomery

  Transformers have redefined sequential recommendation by effectively modeling dynamic user behaviors and long-range dependencies. However, they remain inherently inefficient: standard architectures operate at a fixed rate, allocating comparable computation to every item in a user’s history regardless of its information content. This leads to prohibitive computational overhead on long sequences and increased sensitivity to behavioral noise. To address this, practitioners often resort to lossy sequence compression, staged modeling, or truncation. This limits the model’s ability to leverage the full context of long histories during inference. Inspired by the recent success of Byte Latent Transformers, we propose DP-Rec, a dynamic latent patching architecture for recommendation. DP-Rec shifts from item-level modeling to patch-level modeling by segmenting interaction sequences using contrastive entropy surprise to identify informative behavioral boundaries. A lightweight patch encoder compresses these temporally contextualized segments into a reduced set of dynamic latent behavior vectors, which are then processed by a larger latent transformer and decoded for next-item prediction. Extensive experiments show that, under constrained computational budgets, DP-Rec scales effectively to long sequences and achieves a superior efficiency–accuracy trade-off over both non-compressed and fixed-size compression baselines.

• RESDPGFlow: Decoupled Preference Guided Flow Matching for Cross Domain Sequential Recommendation
  by Xiaoxin Ye, Chengkai Huang, Hongtao Huang, Shoujin Wang and Lina Yao

  Cross-Domain Sequential Recommendation (CDSR) aims to improve next-item prediction by leveraging users’ sequential behaviors across multiple domains. Despite recent progress, existing CDSR models suffer from two fundamental limitations: (1) they often entangle transferable, domain-invariant interests with domain-specific preferences, leading to negative transfer across heterogeneous domains; and (2) they are highly sensitive to noisy interactions such as misclicks, and abrupt domain transitions. Generative models have recently emerged as a promising paradigm for modeling complex preference dynamics and mitigating noise. However, diffusion-based approaches rely on Gaussian initialization and stochastic denoising, resulting in unstable inference. Flow Matching (FM) offers a deterministic and efficient alternative by directly learning preference transport trajectories, yet existing FM-based recommenders are restricted to single-domain settings and fail to account for domain-dependent signals critical in CDSR. To bridge these gaps, we propose DPGFlow, the first Flow Matching framework specifically designed for CDSR. DPGFlow explicitly disentangles user preferences into domain-invariant and domain-specific components and injects them as structured guidance into a domain-aware conditional flow field. This design enables stable and efficient few-step inference, suppresses noise propagation, and facilitates effective knowledge transfer under heterogeneous and noisy behaviors. Extensive experiments on multiple real-world CDSR benchmarks demonstrate that DPGFlow consistently outperforms state-of-the-art baselines, while exhibiting strong robustness under noise, cold-start, and domain-transition scenarios. The data and code are available at here.

• RESDPGR: Dual-Domain Spatiotemporal Generative Retrieval for Intent-Aware Local Life Service Recommendation
  by Lei Shao, Fei Xiong, Meng Wang, Shiqi Tian, Zihan Yang, Ran Li, Xing Dong, Ming Liu and Hao Gu

  Local life service recommendation (LLSR) spans content recommendation and Point-of-Interest (POI) recommendation. On platforms such as Dianping, users browse content (notes, videos, reviews) and interact with POIs (collecting restaurants, planning check-ins) within the same session. The core challenge is that dual domain actions are causally linked: a content click on a food review and a subsequent POI collect are two reflections of the same latent user intent, not merely two separate problems. Yet existing generative retrieval methods treat content and POI signals as independent, missing this shared latent structure. To address these limitations, we propose DPGR (Dual-domain Spatiotemporal Generative Retrieval), a unified dual-domain generative retrieval framework for LLSR. DPGR introduces a Spatiotemporal State-Conditioned Token Modeling mechanism that injects dynamic user context into multiple stages of the encoder, enabling state-dependent reweighting and adaptive preference balancing. Instead of directly generating items, DPGR learns discrete intent codes via quantization of dual-domain behaviors and predicts the top intents for the target session. Each intent code independently retrieves candidates via parallel ANN, achieving diverse coverage with no additional latency. Offline experiments on public and internal datasets show that DPGR outperforms state-of-the-art generative retrieval baselines. Online A/B tests on Dianping demonstrate significant gains in both domains: in the content domain, visit views increase by 1.486% and watch time by 1.015%; in the POI domain, POI clicks increase by 0.527% and POI collects by 5.209% (all < 0.05).

• RESDecoupled Learning and Selection in Slate Recommendation for Privacy and Stability Under Noisy Scores
  by Sam Urmian, Qinyi Liu and Mohammad Khalil

  Many recommender systems do not show users the raw list produced by a learned model. They first score possible items, then apply a repeatable rule layer that removes restricted items, adds variety, enforces constraints, and decides how many items to show. We study what can be said when these two steps are separated explicitly. The aim is not to propose a new recommender, but to understand which privacy, auditability, and stability guarantees follow from this common design pattern. Our main result is a certificate for when the final recommendation list stays unchanged. If the recorded gap between each chosen item and the closest alternative is large enough, then small changes in learned scores cannot change the selected list. This gives a practical way to audit whether a recommendation was robust to noisy scores. The result also explains why mixing a changing model with a fixed reference model can reduce top-list changes under score noise. This stability claim is separate from privacy: if the learned model is trained with a formal privacy guarantee and the rule layer uses only public, fixed, or separately privacy-accounted inputs, then the full system and its audit log inherit that guarantee. If the fixed reference model is trained without privacy protection, the system may still be more stable, but it is not private end-to-end. We test these claims in the settings where they apply. Controlled top-list change tests match the predicted stability pattern, and ranking-change tests on OULAD, MovieLens-25M, and Amazon Musical Instruments show the same stable-reference/noisy-model effect. OULAD and EdNet margin diagnostics find certified cases with no slate changes. Simulated repeated recommendation runs on OULAD and EdNet show that selector rules can bound target drift and make decision changes replayable, while final user-facing utility effects remain mixed. Overall, the paper characterizes what follows from separating learning from repeatable rule-based selection: certifiable stability and scoped privacy claims, not a universally best recommender.

• RESDo We Care About Personalization and Explainability? An Interview Study with News Recommendation Engineers
  by Jasmin Kareem, Siddharth Mehrotra, Martijn Willemsen and Maarten de Rijke

  Research on explainability in recommender systems largely centers on end users, overlooking the perspectives of those who build and maintain these systems and their potential use cases such as model debugging. In this study, we examine how news engineers and related technical stakeholders perceive and implement personalization and explainability in practice. We conducted 15 semi-structured interviews across nine news organizations, spanning diverse regions in both public and private sectors, to investigate the challenges and motivations shaping their approaches. Our findings reveal that personalization is not always a straightforward or desirable choice for news organizations, as concerns around user tracking, editorial control, and resource constraints often limit its adoption. Even among organizations implementing personalized news recommender systems in production, explainability is rarely prioritized, with day-to-day operational demands frequently taking precedence over longer-term transparency goals. Definitions of explainability vary widely across organizations, though some demonstrate promising internal practices and visualization tools that facilitate communication between engineering teams and newsrooms. Based on our analysis, we provide actionable and practical guidelines for news engineers and researchers on how to adopt explainability methods within a news personalization pipeline.

• RESDo We Really Need LLMs to Augment All? A Selective Augmentation Framework with Lightweight Language Models for Multimodal CTR Prediction
  by Ziyun Chen, Yuhan Wang, Honghao Li, Mengzi Tang, Qing Xie and Yongjian Liu

  In recent years, multimodal models and large language models (LLMs) have been increasingly applied to click-through rate (CTR) prediction, owing to their ability to extract recommendation-relevant information from raw content and thereby alleviate the long-standing issue of collaborative signal sparsity. Existing paradigms typically either feed continuous embeddings from pretrained multimodal encoders into CTR models, or employ LLMs for unified knowledge augmentation. While these approaches have demonstrated promising performance, when and why their gains emerge remains unclear, and the inference cost of large LLMs introduces severe latency and scalability challenges in large-scale deployment. In this work, we empirically show that the gains of LLM-augmented CTR are highly item-dependent: uniformly applying stronger augmentation to all items is often inefficient, while the main benefits concentrate on a subset of items with more complex interaction patterns. Motivated by this observation, we propose A Selective Augmentation Framework with Lightweight Language Models for Multimodal CTR Prediction, dubbed SALM, which revisits the data augmentation paradigm and replaces indiscriminate full-data augmentation with a difficulty-aware selective strategy targeting necessary items. Extensive experiments on real-world multimodal datasets, across multiple CTR backbones and augmentation baselines, demonstrate that SALM consistently improves AUC and reduces LogLoss, while simultaneously reducing LLM-based augmentation cost.

• RESEmpowering Cross-Domain Sequential Recommendation with Hybrid Tokenization and Serial-Parallel Decoding
  by Yuxuan Hu, Yuhao Wang, Tianbo Huang, Chao Zhang, Ziwei Liu, Lihua Zhang and Xiangyu Zhao

  Cross-domain sequential recommendation (CDSR) aims to model users’ dynamic interest transitions and sequential patterns across multiple domains. Recently, generative recommendation (GR) has emerged, which first learns semantic identifiers (SIDs) using semantic information of items and models the recommendation task as autoregressive generation. However, it faces with two critical issues: 1) ignoring collaborative correlations across different domains in tokenization step and 2) adopting inefficient decoding strategies like beam search in generation step, which hinders GR’s application in real-time services. To address these issues, we propose GenCDSR, an effective and efficient generative framework for CDSR. Specifically, GenCDSR learns domain-aware SIDs through a cross-domain hybrid tokenization mechanism, which jointly incorporates domain-shared and domain-specific codebooks to capture both cross-domain commonalities and distinctions. Furthermore, we design a serial-parallel decoding strategy that partially parallelizes cross-domain generation while preserving generation consistency, thus significantly reducing inference latency. Experimental results on three public datasets validate that GenCDSR achieves a 1.5% improvement in accuracy and an 85.1% reduction in inference latency on average compared to SOTA baselines. The implementation code and datasets are available online: https://anonymous.4open.science/r/GenCDSR.

• RESExplaining Last-Item Reliance in Causal Self-Attention for Sequential Recommendation via Residual Dominance
  by Keito Kozaki, Keigo Sakurai, Ren Togo, Takahiro Ogawa and Miki Haseyama

  Transformer-based sequential recommenders with causal self-attention are known to rely heavily on the most recent interaction at inference time, but the structural origin of this behavior remains unclear. We study this problem by combining prediction-time diagnostics with norm-based analysis of the full attention block. First, we show that SASRec-style models exhibit highly localized last-item reliance. We then find that, although self-attention aggregates contextual information, the residual pathway has a dominant influence on the final representation and sharply reduces the expression of preceding context, yielding what we term residual dominance. To probe this interpretation, we use inference-time residual scaling as a controlled diagnostic intervention. Changing the residual strength induces a monotonic trade-off between structural mixing and last-item reliance, and reveals that some correct predictive signals already exist beyond the final position but are weakly expressed under standard inference. Overall, our results provide structural evidence that extreme last-item reliance in causal self-attention recommenders is closely tied to the dominance of the residual pathway at inference time. The code is available at: https://anonymous.4open.science/r/Residual-1BC6.

• RESFASC: A Feature Aspect-Level Sentiment Consistency Framework for Explainable Recommendation Evaluation
  by Chenfu Yu, Qinglin Huang, Xiaoxuan Shen, Qian Wan, Zhicheng Dai, Jianwen Sun and Ruxia Liang

  Recent research in explainable recommendation commonly uses natural language explanations to improve transparency and user trust. However, reliably evaluating whether explanations are semantically faithful to users’ multi-dimensional preferences remains challenging. Existing methods mainly rely on text similarity metrics (e.g., BLEU, ROUGE) or shallow feature matching, which are insufficient for assessing whether explanations accurately reflect user preferences, especially in multi-aspect settings. To address these limitations, we propose Feature Aspect-Level Sentiment Consistency (FASC), a framework that quantifies semantic consistency between generated explanations and user-authored reference explanations through aspect coverage and sentiment polarity. FASC uses LLMs as auxiliary tools to extract structured aspect–sentiment units from explanation texts, enabling reproducible metrics for aspect coverage, correctness, and sentiment alignment. To validate the framework, we re-annotated several widely used explainable recommendation datasets to construct benchmarks with fine-grained, aspect-level sentiment labels. We further conducted human studies using pairwise comparisons, showing that FASC aligns more closely with human judgments than traditional metrics. Experimental results indicate that FASC can distinguish subtle differences in semantic faithfulness across models. To support future research, we release all annotated datasets and human evaluation results at https://anonymous.4open.science/r/FASC-F54E

• RESFiCoRec: Fine-Grained Contrastive Learning with Dual Aggregation for Sequential Recommendation
  by Shun Zhang, Ziqiang Yin, Runsen Zhang and Junliang Pan

  Sequential recommendation methods integrated with contrastive learning have been proven effective in addressing the data sparsity issue. However, most contrastive learning schemes directly perform random data augmentation on original sequences, which struggles to capture fine-grained features in users’ historical interaction sequences. Meanwhile, these augmentation methods lack semantic consistency. Additionally, most approaches employ a single aggregation strategy for user representation, making it difficult to comprehensively characterize user preferences. To tackle these issues, we propose a Fine-Grained Contrastive Learning with Dual Aggregation approach for Sequential Recommendation (FiCoRec). Specifically, we design four tailored data augmentation methods on user embedding sequences to ensure semantic consistency and adaptability, and construct rich self-supervised signals, thereby enabling fine-grained contrastive learning. Furthermore, we design a Dual Aggregation module to capture the Tail Aggregation features and Global Aggregation features of sequences, which facilitates the comprehensive learning of users’ short-term key interests and long-term global preferences. Extensive experiments conducted on four public datasets demonstrate that FiCoRec achieves superior performance compared with existing baseline models, with up to 45.93% increase on Mean Reciprocal Rank (MRR).

• RESFrom Constraint to Control: Modeling Expected Fairness in Ranking Systems
  by Tristan Cladière, Antoine Gourru, Bissan Audeh and Christine Largeron

  Modern learning to rank systems have achieved remarkable performance across a wide range of applications. However, they may also exhibit disparities in exposure, raising concerns about fairness, especially in sensitive domains such as healthcare, judicial decision-making, and recruitment, where biased rankings may have critical societal consequences. A common approach to mitigate such issues is to incorporate fairness constraints or regularization terms into the training objective. Yet, this provides limited insight into how these constraints influence the final level of bias, and consequently require costly hyperparameter tuning to reach a desired fairness outcome. In this work, we study fairness in ranking by leveraging threshold-based constraints on disparate exposure, which, under a distributional approximation, induce a predictable transformation of the exposure distribution. We derive a closed-form expression for the expected disparate exposure as a function of the threshold, and introduce an anchored formulation that accounts for practical optimization limits. This formulation enables practitioners to directly select a threshold that achieves a desired fairness target, eliminating the need for extensive hyperparameter tuning. An experimental evaluation on standard learning to rank benchmarks confirms that the proposed model closely matches empirical behavior. These results demonstrate that fairness can be explicitly modeled, predicted and controlled, providing novel and sound approach to tuning fairness in ranking systems. Finally, we disclose our source code for full reproducibility.

• RESGSPRec: On Improving Item Representations in Graph Signal Processing for Collaborative Filtering
  by Ahmad Bin Rabiah and Julian McAuley

  Graph-based collaborative filtering methods act as low-pass filters in the spectral domain and discard the intermediate-frequency components where community-level user preferences reside. Existing GSP-based methods address this through increasingly sophisticated filter designs, yet derive item representations from the user-item interaction matrix alone. The interaction matrix captures which items each user interacted with, but not which items users interacted with close together in their interaction ordering. We propose GSPRec, a graph spectral collaborative filtering framework that produces richer item spectral representations by incorporating item-item proximity derived from user interaction ordering before spectral filtering. GSPRec derives item-item edges from user interaction ordering via multi-hop diffusion and incorporates them into the graph topology. The resulting Laplacian exposes intermediate-frequency structure that a Gaussian bandpass filter selectively amplifies. A low-pass filter retains broad popularity trends. Extensive experiments on four real-world datasets show that GSPRec outperforms all GSP-based and GCN-based CF baselines, with average improvements of 5.12% in NDCG@10. Ablation studies establish that graph construction and filter design are coupled: incorporating item-item proximity without the bandpass filter falls below all GSP baselines, while bandpass filtering without item-item proximity already surpasses them.

• RESGTP: Mitigating Popularity Bias in PLM-based Sequential Recommendation via Group-Aware Token Pruning
  by Ruilin Yuan, Dugang Liu, Hao Chen and Zhong Ming

  Incorporating item textual information with pretrained language models (PLMs) still incurs popularity bias at the semantic representation level. Although existing studies have explored debiasing at the token level, most of them treat tokens in a uniform manner based on item inputs, without finely distinguishing the distributional differences between unique and shared tokens in popular and tail items. Consequently, it remains difficult to characterize the heterogeneous impact of different tokens on popularity bias. To address this issue, we first conduct a validation study to investigate token distribution patterns in popular and tail items. Our findings reveal that unique tokens are more likely to introduce bias than shared tokens, primarily driven by the distributional discrepancies between popular-unique and tail-unique tokens. Based on this observation, we find that strategically pruning biased tokens can enhance item exposure fairness without sacrificing recommendation accuracy. Accordingly, we propose GTP, a group-aware token pruning framework with an adaptive selection network. GTP explicitly categorizes tokens into popular-unique, tail-unique, and shared groups, and learns differentiated retention strategies for different types of unique tokens, thereby effectively mitigating popularity bias induced by unique tokens. Extensive experiments on three real-world datasets demonstrate that GTP consistently alleviates popularity bias while maintaining recommendation performance.

• RESGateBoxGCN: Hard-Soft Gated Box Embeddings with Graph Convolution for Recommendation
  by Fan Mo, Takashi Wada, Rongqin Chen, Chongxian Chen, Xin Fan, Tianwei Chen and Hayato Yamana

  This paper proposes GateBoxGCN, a box embedding framework that relaxes the strict positivity constraint on offsets via a hard-forward/soft-backward gating mechanism, improving recommendation performance. Box embeddings have been explored as a technique to model user preferences via high-dimensional boxes defined by centers and offsets. However, existing box-based methods restrict all offset dimensions to be positive to ensure valid box geometry, limiting the model’s flexibility and expressiveness. To address this limitation, we relax the constraint to allow offsets to be negative values. We then use negative offset dimensions to explicitly model unevaluable dimensions, such as those arising from unobserved user preferences or noise. During inference, we exclude unevaluable dimensions and calculate user-item preference scores by using only the robust ones. To handle unevaluable dimensions, we further introduce a hard-forward/soft-backward gating mechanism, where unevaluable dimensions are filtered out by the hard gate during forward propagation while the soft gate provides gradients to these dimensions during backpropagation, enabling end-to-end learning of the gating mechanism and user/item box representations. Experimental results on real-world datasets confirm the effectiveness of our method over state-of-the-art baselines.

• RESGeneralized Position-Based Model: Rethinking Position Weights in Ranking Off-Policy Evaluation
  by Norman Knyazev, Vito Bellini, Huseyin Yurtseven and Ben London

  Off-policy evaluation (OPE) estimates the performance of new recommendation policies using logged data, thus enabling fast, safe and inexpensive iteration prior to costly A/B tests. To evaluate ranking policies, existing OPE estimators all make structural assumptions about user behavior, leading to a spectrum of trade-offs between bias and variance. The recently proposed INTERPOL estimator navigates these trade-offs through a window system that defines how clicks at different positions are combined. However, this approach has two key limitations: the window configuration must be specified a priori, which can impede practical use, and all positions within a window are weighted uniformly, regardless of their relative utility, potentially limiting accuracy. To address these gaps, we introduce Generalized PBM (GPBM), an estimator that learns position-specific weights by minimizing an approximate upper bound on the estimation error. GPBM retains the unbiasedness guarantees of INTERPOL while eliminating manual hyperparameter tuning. Our experiments demonstrate that GPBM provides more accurate and robust estimates across a wide range of position bias misestimation levels, logging policies, and dataset sizes.

• RESGenerating Personalized Images for Sparse-Interaction Users with Uncertainty-Aware Retrieval and Dense Knowledge Guidance
  by Yuting Zhang, Ying Sun, Dazhong Shen, Ziwei Xie, Feng Liu, Changwang Zhang, Xiang Liu, Jun Wang and Hui Xiong

  Personalized image generation aims to synthesize target images tailored to individual preferences based on users’ historical interaction data. Existing methods typically inject features from historical interaction records to guide personalized generation. However, such methods encounter two critical challenges when serving sparse-interaction users: (1) Preference Misalignment: Sparse interactions tend to lack target-semantic preference information, causing the direct injection of interaction features to misalign with users’ true target preferences. (2) Lack of Reliable Supervision: Mining preferences from sparse interactions requires sufficient supervision signals, yet personalized generation with sparse data inherently lacks direct feedback or ample preference signals for generated outputs. To this end, we propose Uncertainty-aware retrieval with Dense guidance for Sparse personalized Image Generation (UDSIG). For preference alignment, we first retrieve reference images that exhibit low-uncertainty matching with user sparse preferences from the entire dataset. For reliable supervision, we propose a dense-to-sparse scheme that incorporates a reward model derived from active users’ dense interaction data to drive personalized generation in sparse scenarios. Extensive experiments and human evaluations across three public datasets confirm the superiority of our model, along with its strong generalization to dense interaction scenarios.

• RESGive the Long-tail More SPACE: Promoting Provider Fairness in Next POI Recommendation
  by Anran Zhang, Jiaqi Jiang, Jiahui Jin and Yuhan Zhao

  Next point-of-interest (POI) recommendation predicts users’ future destinations from historical mobility sequences and has become a key component of location-based services. However, mainstream models often concentrate exposure on a small set of popular POIs, leaving long-tail merchants systematically under-exposed. While provider fairness has recently attracted increasing attention, directly applying existing provider-fairness techniques to POI recommendation is problematic: (i) users face execution constraints; and (ii) POIs face resource supply constraints. These coupled constraints render provider fairness in POI recommendation a fundamentally different—and more challenging—problem than in purely digital settings. To address this, we propose SPACE (Supply- and Physics-Aware Conditional Embedding generation), a model-agnostic framework that improves long-tail POI exposure via virtual user generation under explicit feasibility and supply control. SPACE consists of three stages: (1) community inference to capture heterogeneous user execution constraints; (2) unbalanced optimal-transport allocation to decide how many virtual users each tail POI should receive from which communities under POI-specific supply budgets; and (3) constraint-guided latent diffusion to generate POI-conditional, community-consistent virtual user embeddings. The generated user–POI pairs can be seamlessly used to train existing recommenders without modifying their architectures. Extensive experiments on three real-world datasets demonstrate that SPACE substantially improves provider fairness while maintaining—and often improving—recommendation accuracy across multiple backbone models. Our code is publicly available at https://anonymous.4open.science/r/anonym046A/.

• RESHamiltonian Spectral-Temporal Dissipative Dynamics for Sequential Recommendation
  by Shuiying Liao and P. Y. Mok

  Sequential recommendation requires understanding how user preferences evolve over time, yet most existing models treat such evolution as a first order process where the next state depends solely on the current latent representation. Nevertheless, real user behavior often exhibits richer dynamics, including inertia, periodicity, and sudden shifts that cannot be fully captured by these first order assumptions. Motivated by these behavioral characteristics, we reconceptualize sequential recommendation through the lens of second order dynamical systems and introduce the Hamiltonian Spectral Recommender (HSR), a novel framework that models user interest trajectories in a dissipative Hamiltonian phase space. HSR decomposes user interest evolution into position (long term preference) and momentum (short term tendency) components, and uses a spectral symplectic integrator to propagate these dynamics efficiently in the frequency domain. A learnable dissipation mechanism further captures natural interest decay, while a short impulse refinement module models abrupt behavioral fluctuations commonly observed in sparse interaction logs. This design jointly accounts for global periodic patterns, inertial evolution, and localized shocks — three phenomena that are underrepresented in existing sequential models. Extensive experiments on three benchmark datasets demonstrate that HSR consistently outperforms state-of-the-art Transformer-based and state space model based recommenders. Ablation studies further verify the necessity of each component, including spectral propagation, second order coupling, and impulse refinement. This study highlights the value of incorporating dynamical systems perspectives into sequential recommendation, offering an effective alternative to prevailing first order modeling approaches.

• RESHierarchical Semantic Tokenization for Generative Recommendation
  by Tianxin Wei, Xuying Ning, Xuxing Chen, Ruizhong Qiu, Yupeng Hou, Yan Xie, Shuang Yang, Zhigang Hua and Jingrui He

  Generative recommendation models next-item prediction as autoregressive generation over tokenized user histories, where each item is represented as a sequence of discrete tokens. However, existing methods typically construct these tokens by compressing heterogeneous item attributes, such as ID, category, title, and description, into a single latent representation before quantization, which obscures the hierarchical structure of item semantics and limits their ability to capture how user preferences evolve from broad interests to specific choices during web interactions. To address this issue, we propose NAME, a generative recommendation framework that explicitly incorporates Coarse-to-Fine semantic structure into both item tokenization and decoding. Specifically, NAME organizes item information into multiple semantic levels, spanning high-level categories, fine-grained textual content, and collaborative signals. Building on this design, we introduce the CoFiRec Tokenizer, which tokenizes each semantic level independently while preserving their structural order, thereby better reflecting how users refine their preferences and enabling more structured generation. During autoregressive decoding, the language model generates item tokens progressively from coarse to fine, allowing the recommendation process to better capture the natural refinement of user intent. Extensive experiments on multiple public benchmarks and backbone models demonstrate that NAME consistently outperforms existing baselines, and our theoretical analysis further shows that structured hierarchical tokenization reduces the expected dissimilarity between generated items and ground-truth targets.

• RESImproving Rare Medication Recommendation with Counterfactual Data Augmentation and Large Language Models
  by Shinhwan Kang, Soo Yong Lee, Jaewon Kim, Kijung Shin and Buru Chang

  AI-based medication recommendation systems have attracted substantial attention due to their potential to enhance patient safety and therapeutic outcomes. Despite the clinical importance of accurately recommending rarely prescribed medications (rare-meds), we observe that most existing methods show significantly lower predictive performance for rare-meds. We attribute this issue to two intrinsic limitations: (a) the inherent scarcity of data for rare-meds and (b) limited consideration of co-recommended medications. To address these limitations, we propose GenRxR, a novel framework based on large language models (LLMs). GenRxR leverages the medical knowledge and clinical reasoning capability of LLMs to generate counterfactual medical data, mitigating the data scarcity issue for rare-meds. It also integrates an LLM into the medication recommendation process to model relationships among co-recommended medications. To further enhance the clinical reasoning, we introduce an instruction tuning step that aligns the LLM’s capability with the recommendation task, enabling better handling of clinical context, including rare-meds cases. In our experiments, we show that GenRxR outperforms 14 (including 5 LLM-based) baselines in most cases. Specifically, it achieves up to 30.9% higher predictive performance for rare-meds than the strongest baseline.

• RESInformation-Aware Long Sequence Compression for Sequential Recommendation
  by Wooseung Kang, Minje Kim, Suwon Lee, Gun-Woo Kim and Sang-Min Choi

  Sequential recommendation (SR) aims to predict a user’s next interaction by modeling temporal dependencies in historical behavior sequences. However, modeling long sequences introduces two challenges: longer histories often include noisy interactions irrelevant to a user’s core interests, and increasing sequence length substantially raises computational cost while often degrading prediction accuracy due to noise accumulation. We present RDSR, a Rate-Distortion-based Sequential Recommendation framework grounded in a task-oriented rate-utility view. Instead of directly modeling full-length sequences, RDSR combines fixed-capacity token selection with VIB-based latent compression to retain task-relevant information under explicit rate control. This suppresses irrelevant interactions while preserving essential preference signals, effectively reducing sequence length and computational overhead. Extensive experiments show that RDSR improves the performance–efficiency trade-off across attention-, MLP-, and SSM-based SR backbones, with performance gains depending on backbone inductive bias. Our code and supplementary material are available at https://anonymous.4open.science/r/Recsys_RDSR-1D6C/

• RESInverse Theory of Mind Modeling for Content Recommendation: From Web Browsing to Dynamic Intelligent Interfaces
  by Mengyu Chen, Feiyu Lu, Chun-Fu Chen, Lucas Vinh Tran and Jay Katukuri

  Modern recommender systems treat observed actions as reliable proxies for user preferences, yet interactions often reflect exploration or comparison rather than stable preference expression. As interfaces evolve from static layouts toward generative UIs and immersive extended reality (XR), the need for deeper, modality-agnostic user understanding grows: these adaptive environments must decide not only what to present but where, when, how prominently, and most importantly why a user acts. We propose an Inverse Theory of Mind (IToM) pipeline that reasons backward from observed interactions to infer the beliefs, preferences, and decision-making traits that explain behavior. The pipeline reconstructs each user’s decision context, including what was chosen and what alternatives were available, applies LLM-driven counterfactual reasoning to produce evidence-grounded natural-language belief statements, and synthesizes these beliefs through multi-hypothesis abductive inference into a structured user persona. We evaluate on the OPeRA dataset against ground-truth personality assessments, attitudinal surveys, and interview-based personas across four tasks: next action prediction, shopping category prediction, Big Five personality inference, and shopping attitude alignment. Results show that inferred personas match or exceed ground-truth personas and that multi-hypothesis reasoning is useful for personality prediction. We further demonstrate cross-modal transferability with a persona-driven spatial banking application on VisionOS.

• RESLearning Sparse Representations of Multimodal Content for Enhanced Cold Item Recommendation
  by Gregor Meehan and Johan Pauwels

  The scale and rapid growth of item catalogs in modern digital platforms present significant challenges to recommender system (RS) practitioners. Most RSs use embedding similarity to predict user-item preferences, but storage and low-latency retrieval of these embeddings is challenging in industry-scale catalogs. Furthermore, newly added items do not have corresponding embeddings and cannot be recommended effectively; previous works often tackle this item cold-start problem by generating cold item representations from auxiliary content, such as images or descriptive text, so that user preferences can be predicted without historical interactions. In this paper, we argue that sparse embeddings have notable advantages over standard dense vectors in this content-based cold-start paradigm. We describe how existing cold-start training regimes can be adapted for sparse representation learning, and build on insights from linear attention to design a pre-sparsification activation technique that induces sharpness and denoising effects in learned item-item similarities. We show that the resulting sparse embeddings achieve significant improvements in cold-start recommendation accuracy over dense embeddings at considerably lower storage costs, especially for users with multiple interests. Through comprehensive experiments on four multimodal RS datasets, we also demonstrate the interpretability of sparse content embeddings and their robustness in the trade-off between size and accuracy.

• RESMODE: Mutual Optimality in Direct Effects of Reciprocal Recommendations in Matching Markets
  by Yoji Tomita

  Matching platforms such as job posting services and online dating platforms have become widely used over the past decade. For a matching platform to be successful, it is crucial to design appropriate reciprocal recommendation systems (RRSs) that consider the preferences of both sides of users (job candidates and employers) and prevent opportunities from being concentrated too heavily on a few popular users. However, prioritizing concentration mitigation too much can lead to recommending undesirable results to some individual users, resulting in their dissatisfaction. In this paper, we formulate the concept of optimality of direct effects of the recommendation list for an individual user, given the recommendations to other users. Furthermore, we propose a novel method, MODE, that computes mutually optimal recommendations in direct effects. Experiments with synthetic and real-world data demonstrate that MODE surpasses other existing methods in terms of mutual optimality of direct effects, exhibits faster processing speeds, and enables a higher expected number of matches.

• RESMaskPoison: Intent-Guided Poisoning Attacks on Sequential Recommendation via Masked Discrete Diffusion
  by Han Zhou, Hongxu Ma, Hui Fang, Jiayu Xu and Zhu Sun

  Sequential recommendation (SR) systems are widely deployed across modern online platforms and have been shown to be vulnerable to poisoning attacks. Such attacks inject fabricated user sequences into training data to promote target items. Existing methods achieve stealthiness by enforcing surface-level similarity to genuine data, including matching item frequencies, local transition patterns, and co-occurrence statistics. This assumption holds in idealized, homogeneous settings where user behaviors are narrow and repetitive. In real-world platforms, however, users exhibit rich and context-dependent behavioral intents. In such heterogeneous environments, surface-level mimicry fails to preserve the logical coherence of user intent, causing poisoned sequences to be detectable. We propose MaskPoison, an intent-guided poisoning framework that addresses this fundamental gap. Our method extracts intent anchors from real user sequences containing the target item. A masked discrete diffusion model then synthesizes poisoned sequences conditioned on these anchors as hard semantic constraints, ensuring alignment with the behavioral manifolds of genuine users. Extensive experiments demonstrate that MaskPoison outperforms existing attacks in both attack effectiveness and stealthiness, across homogeneous and heterogeneous recommendation scenarios alike. Our code is available at https://anonymous.4open.science/r/MaskPoison-Code-6678/.

• RESMembership Inference Attacks on In-Context Learning Recommendation
  by Jiajie He, Min-Chun Chen, Xintong Chen, Xinyang Fang, Yuechun Gu and Keke Chen

  Large language models (LLMs) based recommender systems (RecSys) can adapt flexibly across different domains. It uses in-context learning (ICL), i.e., prompts, including sensitive historical user-specific item interactions, to customize the recommendation functions. However, no study has examined whether such private information may be exposed by novel privacy attacks. We design two membership inference attacks (MIAs): ItemMem, and RecInertia, aiming to identify whether system prompts contain the victim’s information. We have carefully evaluated them on the latest open-source LLMs and three well-known RecSys datasets. The results confirm that the MIA threat to LLM RecSys is realistic and can be more sophisticated than prompt extraction. They utilize the unique prompt structures in ICL RecSys and cannot be easily mitigated with existing defense methods on prompt extraction.

• RESNot All Branches Are Equal: Adaptive Semantic ID Construction for Generative Recommendation
  by Guy Hadad, Haggai Roitman and Bracha Shapira

  Generative recommendation models represent items as sequences of discrete tokens known as semantic IDs. Existing approaches derive these identifiers using residual or hierarchical vector quantization, assuming that tokens carry independent semantic meaning across quantization levels. Revisiting this assumption, we argue that semantic IDs should behave as hierarchical paths where each token’s meaning is conditioned on its prefix. This highlights a mismatch between conventional fixed-codebook quantization and the hierarchical structure implicitly learned by generative recommenders. Motivated by this insight, we propose Adaptive Silhouette Tree (AST), a top-down divisive clustering method that constructs semantic IDs as paths within a tree structure. Unlike prior methods with fixed branching factors, AST dynamically determines the number of branches at each node by maximizing the silhouette coefficient, enabling finer partitions in dense regions and compact groupings in homogeneous ones. We further introduce constrained variants that enforce minimum probability mass during splitting to improve supervision for long-tail items. Extensive experiments on five Amazon Review datasets of varying sizes and domains show that AST significantly outperforms strong baselines. These results demonstrate that aligning semantic ID construction with the inherent hierarchical structure of generative recommenders leads to more effective recommendations.

• RESOn Reranking Space for Multi-Tenant Retrieval with Adapted Queries
  by Jun Woo Chung and Weijie Zhao

  Multi-tenant dense retrieval systems increasingly employ shared compressed indexes where individual tenants adapt embeddings via fine-tuning (e.g., LoRA). While query-side projection adapters bridge the resulting embedding mismatch, a critical design choice remains for the optional reranking stage: should distances be computed in the original index space (source-space) or the adapted query space (target-space)? Contrary to the intuition that the calibrated target-space should perform better, we find the opposite to be true. Across 23 LoRA-adapted dataset-tenant pairs, four adapter architectures, and four index configurations (ranging from PQ-16 to HNSW), source-space reranking consistently outperforms target-space reranking, improving nDCG@10 by more than 10 percentage points in some cases. We further evaluate distance blending between these signals, finding that it provides robust gains on coarse indexes (e.g., PQ-16) when the reverse adapter is structurally sound, while adding minimal latency. Our results offer a straightforward heuristic for multi-tenant platforms: maintain the shared index, project queries using existing adapters, rerank in the source-space, and apply distance blending as latency permits when working with low-precision indexes.

• RESOn the Convergent Validity of Offline Evaluation Designs for Recommender Systems
  by Sushobhan Parajuli, Samira Vaez Barenji and Michael Ekstrand

  Offline evaluation on historical interaction logs is the most common evaluation methodology for recommender systems. However, such evaluations depend on sparse, incomplete or biased data which raises concerns about whether commonly used evaluation setups reliably reflect true user preferences. In this work, we study how offline evaluation design choices affect the validity of recommender system comparisons. We evaluate a set of recommendation models across many evaluation configurations that vary key factors including data filtering thresholds, feedback binarization versus graded relevance, candidate set construction, train-test splitting strategies, and evaluation metrics. To assess the validity of these configurations, we measure the correlation between model rankings obtained from conventional train-test splits on sparse interaction data and rankings from evaluations based on dense ground-truth relevance judgments. We use this agreement as an evidence of their validity with respect to true user preferences. Using KuaiRec and extended MovieLens-32M datasets that provide such ground-truth data, we analyze which evaluation setups produce results that better align with ground-truth performance.

• RESProbabilistic Residual Learning for Online Recommendations
  by Wenyuan Wang, Yusong Zhao, Zihao Xu, Hengyi Wang, Qi Xu, Zhigang Hua, Yan Xie, Yi Wang, Zihao Zhao, Bo Long, Chengzhi Mao, Shuang Yang, Hengguan Huang and Hao Wang

  Modern recommender systems are typically based on deep learning (DL) models, where a dense encoder learns representations of users and items. As a result, these systems often suffer from the black-box nature and computational complexity of the underlying models, making it difficult to systematically enhance their recommendation capabilities. To address this problem, we propose Probabilistic Residual Learning (PRL), a causal Bayesian recommendation model that models the residual between ground-truth and base predictions, enabling targeted refinement of existing systems. Specifically, PRL (1) divides users into clusters in an unsupervised manner and identifying causal confounders that influence latent variables, (2) learns sub-models for each confounder given the observable variables, and (3) generates recommendations by aggregating the rating residuals under each confounder using do-calculus. Experiments demonstrate that our plug-and-play PRL is compatible with various base DL recommender systems, improving their performance while automatically discovering meaningful user clusters. Auxiliary materials (including the Appendix) are at https://anonymous.4open.science/r/PRL_Appendix-CDD7/PRL_RecSys_Appendix.pdf.

• RESPurifying Interaction Sequences: Topology-Aware Spectral Denoising for Side-Information Integrated Sequential Recommendation
  by Yang Jiao, Chao Yang, Bin Jiang and Junhao Gao

  Side-information Integrated Sequential Recommendation (SISR) enhances preference modeling under sparse interactions by incorporating auxiliary item attributes. However, two core challenges remain. First, real interaction sequences inevitably contain noisy behaviors that distort user preference inference. Second, existing methods often fail to simultaneously preserve ID-driven transition patterns and fully exploit attribute-aware semantics. To address these challenges, we propose TSD-SR. Our approach first represents each fixed-length interaction sequence under a circular-shift topology, then applies Topology-aware Spectral Denoising, followed by a Decoupled Dual-branch Fusion module. Specifically, Topology-aware Spectral Denoising is performed on two streams: a pure ID stream and an early-fused stream that integrates item ID, attribute, and positional embeddings. This process suppresses accidental interactions and transient fluctuations prior to deep encoding, while preserving informative sequential patterns. Subsequently, a Decoupled Dual-branch Fusion module models ID-centric transition dependencies and attribute-enriched semantic context in parallel, capturing collaborative signals and side-information semantics while alleviating information invasion. Experiments on four datasets from two real-world benchmarks, namely Yelp and three domains from Amazon Reviews, demonstrate that TSD-SR consistently outperforms state-of-the-art sequential recommenders and side-information-aware baselines, while exhibiting robust performance in long-tail and noisy settings. Code: https://anonymous.4open.science/r/TSD-SR-2374

• RESRecRec: Latent Interests Recursive Reasoning for Sequential Recommendation
  by Wenhao Deng, Junchen Fu, Hanwen Du, Alexandros Karatzoglou, Ioannis Arapakis, Hangjun Guo, Kaiwen Zheng, Yongxin Ni and Joemon Jose

  Sequential recommender systems rely on a single forward pass to encode user interaction histories and predict the next item. Increasing inference-time computation through latent reasoning, with the model proceeding step by step before the final prediction, has been recently explored in sequential recommendation with promising results. However, how to structure the reasoning process for sequential recommendation remains an open question. Existing approaches couple reasoning and prediction in a single d-dimensional state, limiting reasoning depth and often relying on multi-stage pipelines with reinforcement learning. We propose RecRec (Recursive Reasoning for Recommendation), an RL-free framework that decouples reasoning from prediction, overcoming the fixed d-dimensional state bottleneck of prior methods. RecRec consists of a Context Compressor and a Recursive Reasoner, trained in two simple supervised stages. The Context Compressor distills the backbone’s hidden states into a small set of latent interests, with an Interest Diversity Regularizer encouraging each interest to capture a distinct aspect of user behavior. The Recursive Reasoner then refines these interests by reasoning in a separate intermediate latent space. Deep supervision lets the reasoning depth be freely adjusted at inference without retraining. On four real-world datasets, RecRec outperforms state-of-the-art reasoning-enhanced methods, and on three of four datasets, gains extend past the training-time depth. Our findings point to a decoupled, multi-vector recipe that unleashes latent reasoning from the single-state bottleneck of prior methods, suggesting reasoning-state structure as a design axis to explore further in sequential recommendation.

• RESReducing Perceived Polarization through Affect-Balanced News Reframing
  by Jia-Hua Jeng, Alain D. Starke, David Elsweiler and Christoph Trattner

  Emotionally charged news headlines can amplify negative reactions and increase perceptions of societal polarization. In this paper, we investigate whether large language models can be used to reframe news headlines in a more affect-balanced manner, combining fear and hope, without undermining engagement or perceived fairness. Across three studies, we analyze emotional framing in news headlines using large-scale interaction data, validate that LLMs can reliably generate fear-hope reframings, and evaluate their effects in a controlled user study (N = 80). Our results show that fear-hope reframed headlines significantly reduce perceived polarization compared to original headlines (roughly a 20-25 decrease relative to the original condition’s mean, substantial given the brief exposure), while reducing negative emotional states such as anger and hostility. At the same time, we find no significant decrease in engagement intentions, perceived fairness, or willingness to pay for news. Emotional effects are not uniform, indicating that LLM-based reframing changes specific emotional responses rather than broadly reducing emotionality. Together, these findings suggest that carefully designed LLM-driven headline reframing can reduce perceived polarization and some negative emotional responses while preserving key engagement-related outcomes.

• RESSAGE: Sequence-level Adaptive Gradient Evolution for Generative Recommendation
  by Yu Xie, Xingkai Ren, Qi Ying, Di Jia and Yao Hu

  Generative recommender systems hold the promise of jointly optimizing accuracy, content diversity, and creator exposure fairness. However, current reinforcement learning–based optimizers such as Gradient-Bounded Policy Optimization (GBPO) exhibit a Symmetric Conservatism failure mode: symmetric update bounds suppress learning from rare positive signals (e.g., cold-start items), static negative-sample constraints fail to prevent diversity collapse under rejection-dominated feedback, and group-normalized multi-objective rewards produce low-resolution training signals. These limitations directly harm user experience by reinforcing information cocoons and reducing new-creator visibility. We propose SAGE (Sequence-level Adaptive Gradient Evolution), a unified optimizer for list-wise generative recommendation. SAGE introduces (i) a geometric-mean importance ratio for sequence-level signal alignment, (ii) asymmetric adaptive bounding—a Positive Boost for cold-start slates and an Entropy-Aware Penalty for low-diversity failures—and (iii) a decoupled multi-objective advantage estimator. On three Amazon Product Reviews datasets and the large-scale RecIF-Bench, SAGE consistently improves top-K accuracy while delivering +89% to +101% cold-start recall recovery and +11% diversity gains relative to GBPO. Beyond-accuracy evaluation confirms that SAGE substantially reduces intra-list similarity and broadens catalog coverage, suggesting that asymmetric, sequence-aware policy optimization is an effective approach to improving both recommendation quality and content ecosystem health.

• RESSPEAR: Selection-aware Personalized End-to-end Adaptive Rewriting and Retrieval for Community Search
  by Wenbin Wu, Yuzhong Wu, Yufan Xu, Kuan Fang, Xing Xu, Cheng Ye and Xiaobin Hu

  Query reformulation bridges user intent and retrieval in e-commerce search, yet production systems optimize rewrite quality and retrieval effectiveness separately, leaving the two stages structurally misaligned. Path-based architectures unify them end-to-end but were designed for personalization, where relevance is not an explicit constraint—search additionally requires the rewrite to remain faithful to the user’s stated query intent. Transplanted directly, these models learn a shortcut we term the generalization-word dominance effect: they favor generic rewrites that score well on paths but drift from query intent. To address this, we propose SPEAR (Selection-aware Personalized End-to-end Adaptive Rewriting and Retrieval), which integrates three components each targeting one failure mode: (1) a dual-embedding backbone with gradient isolation that shields recall-side semantics from being eroded by CTR-driven ranking signals; (2) a multiplicative gating aggregator that lets a rewrite score high only when both its confidence and item relevance are strong, eliminating the generic-rewrite shortcut; (3) a relevance-aligned auxiliary loss that steers the selector toward rewrites faithful to the original query intent, directly enforcing rewrite relevance. Offline evaluation on 100K held-out industrial search sessions shows that the proposed framework improves rewrite semantic similarity@10 by +18.2% and click recall@10 by +99.5% over the production baseline. In online A/B testing, SPEAR achieves +0.259% in query-view CTR and +0.733% in average reading depth, confirming that improved rewrite selection translates into stronger retrieval and deeper user engagement.

• RESSafety-Aware Next-POI Recommendation with Large Language Models
  by Rami Zaboura, Ludovico Boratto and Adir Solomon

  Point of Interest (POI) recommendation has become a core task in location-based services, with modern systems increasingly driven by deep learning models that achieve strong predictive accuracy. Yet, despite these advances, most approaches optimize primarily for relevance, giving limited attention to an equally important real-world factor: user safety. In this study, we propose a safety-aware next-POI recommendation method that leverages a Large Language Model (LLM) to generate predictions informed by both mobility patterns and crime-derived safety signals. By integrating crime statistics with POI data and encoding safety information directly into trajectory prompts, our approach produces recommendations that better reflect real-world risk. Through tailored prompt engineering, we finetune an LLM to incorporate safety considerations, yielding predictions that align with user preferences while prioritizing personal security. Experimental results show that our method substantially improves the safety profile of recommended POIs and surpasses state-of-the-art baselines in overall accuracy.

• RESSimulating Diverse User Behavioral Stereotypes for Evaluating Agentic Conversational Recommenders
  by Alessandro Petruzzelli, Alessandro Francesco Maria Martina, Cataldo Musto, Marco de Gemmis, Pasquale Lops and Giovanni Semeraro

  Agentic Conversational Recommender Systems (ACRSs) are designed to recommend through multi-turn dialogue with users whose needs are not fully formed at the outset. However, their evaluation almost exclusively relies on user simulators that instantiate users with clear, pre-formed needs, reducing the interaction to a retrieval over attributes disclosed in the initial turns of the conversation. This covers only a narrow slice of the behaviors real users exhibit, and assessing the robustness and reliability of these systems requires simulated users that span a wider range. To this end, we introduce a stereotype-conditioned, open-weight user simulator that spans three behavioral stereotypes: Direct, Vague-Proactive, and Vague-Reactive. Benchmarking four state-of-the-art ACRSs across four e-commerce domains with our simulator, three findings emerge. First, under certain stereotypes the user stops contributing new information about the target as turns accumulate while the agent continues to act, a regime previously unobserved, which we name Unproductive Stagnation and formalize via Preference Coverage. Second, a systematic Robustness Gap emerges: as the simulated user shifts from decisive to passive, accuracy collapses while conversations grow longer. Third, accuracy degrades more sharply than Preference Coverage does, decomposing the gap into two separable capabilities current ACRSs lack: elicitation and retrieval, which current evaluation entangles in a single score. Our simulator makes this distinction reportable and gives the field a controllable axis along which elicitation and retrieval can be measured and compared.

• RESStabilizing Stability and Plasticity in Graph-based Continual Recommender System
  by Yixin Chen, Xiangmeng Wang and Qian Li

  Real-world graph-based recommender systems face increasing challenges as interaction graphs evolve continuously, exposing models to persistent out-of-distribution shifts. Continual learning has emerged as a promising paradigm for incremental updates without retraining from scratch. However, existing methods primarily emphasize preserving historical knowledge (i.e., stability) and fail to address the unique challenges of graph structures. We identify fundamental challenges in graph-based continual recommendation. Through empirical analysis, we reveal three key challenges: (i) over-stabilization induced by message passing limits the absorption of new knowledge, i.e., lack of plasticity; (ii) improving plasticity for new items degrades performance on historical items, exposing a stability–plasticity trade-off; and (iii) evolving graph topology weakens the preservation of learned representations, i.e., limited stability. To address these challenges, we propose SSPRec, a graph prompt-based continual learning framework for OOD recommendation that explicitly preserve plasticity, enhance stability, and effectively balance the stability–plasticity trade-off. SSPRec freezes a pre-trained backbone and adapts to evolving graph slices via lightweight prompts and user-specific control. Specifically, we design contextual and temporal prompts to enhance stability, introduce forward-knowledge-guided contrastive objectives to improve plasticity, and develop a preference-shift-aware mechanism to adaptively balance stability and plasticity at the user level. Extensive experiments demonstrate that SSPRec consistently outperforms state-of-the-art methods under evolving graph settings.

• RESTAGRec: Tailness-Aware Gate for Sequential Recommendation under Long-Tailed Distributions
  by Fengying Li, Jiawei Gao, Sen Li and Rongsheng Dong

  Sequential recommendation performance is persistently hindered by the long-tailed distribution of interaction data. Existing dual-view methods can alleviate this issue to some extent, but they typically rely on static fusion strategies and cannot adaptively regulate the contributions of semantic and collaborative signals at the sample level, nor can they transfer such decisions to training-time supervision. We propose TAGRec (Tailness-Aware Gated Recommendation), a unified tailness-aware framework that uses a single gate to coordinate both forward view fusion and backward supervision scheduling. Specifically, TAG estimates a contextual tailness prior from observable histories to produce a sample-level gate, T-MoV uses this gate to dynamically balance semantic and collaborative views, and ASCL preserves the semantic view as a stable anchor while adaptively strengthening asymmetric contrastive supervision on the collaborative branch. Experiments on four public benchmarks show that TAGRec consistently achieves the best or near-best performance, with especially clear gains in tail-heavy settings, while introducing only lightweight overhead and no extra online inference.

• RESTRACE: Targeted Ranking-Aware Counterfactual Explanation for Sequential Recommendation
  by Ungsik Kim, Sang-Min Choi, Gun-Woo Kim and Suwon Lee

  Ranking-constraint counterfactual explanation for sequential recommendation requires query-limited search to decide where to edit and what to substitute—the bottlenecked for query efficiency lies more in how the search space is structured than in the mutation rate alone. We propose TRACE (Targeted Ranking-Aware Counterfactual Explanation), which decomposes the search into three stages under an embedding-accessible, training-free setting: influence-guided position selection, plausibility-aware candidate retrieval, and actual-margin beam pruning. Across five datasets and three recommender architectures, TRACE outperforms the GA-based baseline GECE in validity, query efficiency, edit cost, and likelihood preservation, with the largest gains on bring-in (raising a target item to top-1) and up to an order-of-magnitude reduction in queries on push-out (displacing the current top-1 beyond top-K). Ablations confirm that the gains arise from structuring the search space before evaluation, rather than from increased mutation frequency. Available code: https://anonymous.4open.science/r/recsys-anon-1D31/

• RESTSMOO: Solving Multi-Objective Experimentation with Constrained Thompson Sampling
  by Krishna Chaitanya Kalagarla, Yi Liu, Lin Chai and Wenyang Liu

  Traditional online A/B experimentation limits the number of treatments that can be evaluated concurrently. Bandit-based adaptive experimentation algorithms address this by dynamically reallocating traffic across an order of magnitude more treatments. Yet existing methods involve a fundamental trade-off: single-metric Thompson Sampling-based methods are robust to novelty effects but cannot accommodate multiple launch criteria, while elimination-based multi-objective methods support multiple constraints but risk prematurely removing promising treatments. We introduce TSMOO (Thompson Sampling with Multi-Objective Optimization), a method that bridges this gap by combining multi-metric optimization with continuous learning. Grounded in stochastically constrained best-arm identification, TSMOO extends single-metric Thompson Sampling to multi-objective batch traffic allocation by estimating multi-constraint feasibility at each allocation step, while preserving all treatments throughout exploration. It further incorporates uplift modeling that mitigates temporal effects shared across treatments and control and ensures that the Gaussian distribution assumption holds. In simulations replaying historical real-world experimentation patterns, TSMOO achieves 93-94% success rates in multi-winner settings and 63-66% under novelty effects, outperforming both single-metric and elimination-based baselines.

• RESThe Cost of Continuous Time: Diagnosing Solver Sensitivity in Recommenders and Mitigating It via Training-Free Routing
  by Zixu Li and Sergio Augusto Romaña Ibarra

  Continuous-time recommendation models irregular user behavior naturally, but under heavy-tailed temporal gaps the solver becomes part of the deployment problem. Under an aligned evaluation protocol, we find that explicit integrators (e.g., explicit Euler) exhibit substantial utility degradation on extreme temporal gaps under temporal corruption, a pattern consistent with stiffness-related numerical sensitivity. Implicit solvers mitigate this degradation through fixed-point iterations, but their iterative computational graphs substantially increase tail latency, approaching a two-fold increase in P99 under our fixed software stack. To address this trade-off, we propose Stiffness-Aware Routing (STAR), a training-free, quantile-calibrated inference-time routing policy. Rather than relying on parameterized gating networks, STAR uses offline temporal priors to route only high-risk long-gap requests to an implicit solver while keeping the remaining traffic on the fast explicit path. Experiments on Amazon, Yelp, and ML-1M show that STAR improves the Pareto frontier between robustness and latency under the aligned evaluation protocol. On Amazon, for example, STAR confines utility degradation to 2.52% (compared with 7.80% for Always-Explicit) while capping P99 latency at 18.44ms, thereby avoiding the 24.46ms tail-latency cost of an Always-Implicit baseline. On dense ML-1M, however, Always-Implicit underperforms Always-Explicit, indicating that no single solver is uniformly optimal across temporal regimes. Overall, the results show that deployment-time solver allocation is more effective than uniformly applying a single integration strategy in the ODE-based setting studied here.

• RESThe Utility of LLMs in Recommender Systems Explanation Evaluation
  by Kathrin Wardatzky, Oana Inel, Luca Rossetto and Abraham Bernstein

  Explanations play a crucial role in creating trustworthy recommender systems (RS), yet choosing a good explanation method comes with challenges. Many explanation methods are available, but little guidance exists on which is best for which setting. Existing explanation generation methods often produce abstract outputs that require further formatting to become user-friendly with a seemingly endless pool of options. Running user-based evaluations of all possible options is usually unfeasible, but existing automated evaluation metrics often either assess only the explainer’s abstract output or require comparison with a ground truth, which is generally unavailable. Recent studies have shown that large language models (LLMs) can serve as judges in explanation evaluations, but their reliability has not yet been thoroughly explored. This paper investigates the utility of LLMs in selecting an effective explanation method for a given application. We first explore their ability to generate explanation prototypes given varying information about the RS and the user in the prompts. Specifically, we generate 18 distinct explanation prototypes, which are subsequently evaluated by 14 LLMs of varying sizes across two temperature settings. We compare these against human ratings derived from a user study. Our results show that while LLMs exhibit human-like rating patterns and achieve moderate rank correlation with human raters, their absolute rating agreement is low and varies substantially by model size and evaluation construct. We derive four practical recommendations: keep explanation-generation prompts concise, prefer larger models for evaluation, pre-test evaluation constructs, and audit explanations for factual accuracy, as neither humans nor LLMs reliably detect non-factual content.

• RESThere’s Something About You: Epistemic Recommendation for Latent Interest Discovery
  by Daniel Nemirovsky, Priya Khokher, Adarsh Jois, Marco Zagha and Joaquin Delgado

  How well does a recommender system know you? These systems are typically trained on the silhouette of user activity to predict immediate engagement, such as the next click or stream. Yet this narrow focus may paradoxically expose how fragmented and incomplete the system’s knowledge of the user really is. In contrast to recommending from established user knowledge, in this work, we recommend to enrich it – an approach we term epistemic recommendation. To realize this, we propose EGRec (Epistemic Gain Recommender). For each item it could recommend, EGRec constructs two hypothetical futures (the user engages, or does not) and uses Jensen-Shannon divergence to quantify how much either outcome would enrich the model’s understanding of the user across the full catalog. We explore two variants depending on how outcomes are valued: epistemic gain (EG), which considers what the model would learn from any outcome, and expected epistemic gain (EEG), which weights learning by how likely each outcome is. For practicality, we train a lightweight prediction head on frozen model embeddings to approximate EG in a single forward pass. We formalize our intuition by introducing a dual regret framework analyzing both reward regret and coverage regret, and prove that EGRec achieves bounded reward regret O(T_explore) and bounded coverage regret O(1), while greedy relevance incurs linear coverage regret Ω(T). We evaluate EGRec both as a standalone strategy and as a signal augmenting bandit methods on KuaiRec and MovieLens-1M, against baselines spanning UCB, Thompson Sampling, MMR, and greedy relevance. Our results show that combining EG with bandit exploration (UCB+EG) achieves the best balance of recommendation accuracy and interest coverage, improving alpha-NDCG by 9-19% over the strongest exploratory baseline and calibration by up to 20%. Moreover, we show that EG signal quality depends on the expressiveness of the base model’s user representation, a dependency that serves as a practical diagnostic for when the approach will be most effective. Our work points to a broader principle in recommendation, that the value of recommending an item lies not merely in expected engagement but extends to what the system is poised to learn about the user from the outcome.

• RESTopology-Aware Tokenization for Generative Recommendation
  by Yaokun Liu, Yifan Liu, Zhenrui Yue, Gyuseok Lee, Zelin Li, Ruichen Yao and Dong Wang

  Generative recommendation has emerged as a powerful paradigm by reframing sequential recommendation as an autoregressive generation task. Central to this paradigm is item tokenization, which quantizes continuous item embeddings into discrete semantic IDs for autoregressive item prediction. Despite its importance, we identify a critical yet overlooked issue in the tokenization process: topology distortion. Specifically, we observe that the intrinsic adjacency relationships of items in the continuous embedding space are significantly disrupted after quantization. This topology distortion misleads the model’s perception of item similarity, ultimately bottlenecking the accuracy of generative recommendations. To address this issue, we propose Hierarchical Topology Distillation (HiToD), a topology-aware tokenization framework that preserves item relational structure throughout the quantization hierarchy. Different from the prior monolithic supervision in tokenization, HiToD introduces a multi-level distillation scheme to progressively recover the topology from coarse to fine granularity: 1) Inter-Group Distillation to capture global cluster-wise relations; 2) Intra-Group Distillation to refine local structures within semantic clusters; and 3) Inter-Item Distillation to enforce fine-grained alignment at the individual item level. Extensive experiments on three benchmark datasets demonstrate that HiToD effectively alleviates topology distortion and consistently outperforms state-of-the-art tokenizers, achieving significant performance gains of up to 9.42% in Recall@5. Our code is available at https://anonymous.4open.science/r/HiToD.

• RESTowards welfare-oriented recommendations in activity-travel behavior
  by Ekin Ugurel and Takahiro Yabe

  While mainstream recommender systems (RS) rely on diverse heuristics to rank alternatives, they generally lack a principled account of user welfare (i.e., whether accepting the recommendation will leave the user better off than other alternatives). The problem is particularly acute in activity-based travel behavior, where users incur costs they cannot recoup (i.e., energy, time) regardless of eventual satisfaction. As a result, existing systems may recommend options based on popularity or collaborative filtering, but may still leave users worse off than nearby or self-selected alternatives. We address this gap by introducing a welfare-oriented framework for activity recommendation that evaluates suggestions in terms of net utility, defined as experienced benefit minus travel costs. Specifically, we formalize two operational decision criteria: Positive Utility Probability (PUP) recommends only when the probability of non-negative net utility exceeds a threshold, while Regret Minimization (RM) recommends only when expected regret relative to the user’s best organic alternative falls below a tolerance level. To evaluate these criteria, we develop an agent-based simulation in which heterogeneous synthetic travelers interact with multiple RS over time in a spatial environment with realistic travel costs, congestion, and behavioral feedback loops. This framework enables controlled counterfactual evaluations, and offers a practical foundation for designing RS that treat user welfare as a primary objective rather than an incidental byproduct.

• RESUniRec: A Unified Expressive-Aligned Generative Recommendation Framework for E-commerce
  by Ziliang Wang, Gaoyun Lin, Xuesi Wang, Shaoqiang Liang, Yili Huang, Weijie Bian, Li Zhang, Mingchen Cai, Jian Dong and Guanxing Zhang

  Traditional discriminative recommendation pipelines suffer from objective misalignment and error propagation across stages, motivating a shift toward generative recommendation (GR). However, existing GR methods decode over compact Semantic ID (SID) tokens without access to item-side features, lacking the explicit user–item feature crossing that discriminative models rely on. Combined with the inherent one-to-many nature of recommendation, this absence of item-side signals significantly amplifies generation uncertainty, making the generative paradigm widely regarded as having a lower performance ceiling than its discriminative counterpart. We propose UniRec, a unified expressive-aligned GR framework that unifies the multi-stage pipeline into a single generative model and aligns its expressive power with the discriminative counterpart. Via Bayes’ theorem, we show that any practical gap stems from feature coverage rather than modeling asymmetry, motivating Chain-of-Attribute (CoA), an expressive-alignment mechanism that pre-generates item attributes before decoding SIDs, recovering item-side feature crossing and yielding measurable per-step entropy reduction. Beyond CoA, Capacity-constrained SID enforces exposure-weighted load balancing to suppress token collapse, and Conditional Decoding Context (CDC) injects scenario-conditioned signals to stabilize multi-scenario decoding and Cartesian-product-based structured summaries of generated tokens to reinforce conditional dependence across decoding layers. A joint Reward-Driven Fine-tuning (RFT) and Direct Preference Optimization (DPO) framework further aligns the model with business objectives. Deployed on a large-scale e-commerce platform, online A/B tests confirm significant gains in page-view click-through rate (PVCTR, +5.37%), orders (+4.76%), and gross merchandise volume (GMV, +5.60%).

• RESWhen Do Contrastive Explanations Really Matter in Recommender Systems?
  by Thi Ngoc Trang Tran, Sebastian Lubos, Alexander Felfernig, Mehrdad Rostami, Viet-Man Le and Damian Garber

  Recommender systems often provide explanations to help users understand why items are suggested. Beyond such non-contrastive explanations, systems can also offer contrastive explanations, such as proposing alternative options (“Something different?”) or explaining why a specific item was not recommended (“Why not Item Y?”). This paper investigates when contrastive explanations matter across different contexts. We report on a user study examining two contrastive explanation types across item domains (movies, shopping, and housing) and recommendation algorithms (collaborative, content-based, and constraint-based). Our results show that contrastive explanations yield context-dependent differences compared to baseline explanations on traditional explanation goals, with generally small to moderate effects. Users tend to perceive contrastive explanations as more necessary in higher-involvement domains and in constraint-based recommendation scenarios, where reasoning about trade-offs, alternatives, and excluded options becomes more important, while baseline explanations remain effective for efficient decision-making. Overall, these findings suggest that contrastive explanations should be viewed as complementary mechanisms and selectively deployed based on recommendation contexts.

• RESWhen attention is bounded, structure matters: personalizing recommendation explanations under time constraints
  by Deo Munduku and Elsa Negre

  Recommendation explanations are often consulted in situations where reading time is limited—for example, when users browse options on a mobile device while on the move—and can process only part of the available information. Yet, little is still known about how such explanations should be structured in this kind of context in order to continue supporting decision-making. In explainable recommender systems, personalization has so far focused mainly on explanation content or linguistic form, much less on explanation structure. In this paper, we study explanation structure as a personalization variable under time constraints. To this end, we introduce a user interpretive schema, defined as an explicit representation of the relative importance the user assigns to the different pieces of information relevant to their decision, and we use this schema to structure the explanation by ordering this information according to its importance to the user. From this, we derive a structuring principle: when reading time is limited, the information that matters most to the user should appear as early as possible. We evaluate this principle in a controlled user study involving 663 participants in a restaurant recommendation scenario, where explanation variants differ only in the order in which information is presented. The results show that a structure aligned with the user’s interpretive schema improves the explanation’s ability to support decision-making when time constraints are strong, whereas this effect weakens as more time becomes available. These findings suggest that explanation structure is not merely a matter of presentation, but constitutes a genuine personalization.

• RESZero-Observation User Reactivation with Gap-Driven Dimensional Gating
  by Jiandong Ding, Tianying Liu, Fuyuan Liu, Huijie Qin and Tiandeng Wu

  Sequential recommendation (SR) models excel at capturing continuous user interests, but struggle with user reactivation – scenarios where users return after prolonged absence. We formally define the Zero-Observation Reactivation problem: a user possesses rich pre-gap behavioral history, yet the platform observes strictly zero signals during a macro-gap (delta t spanning months to years), causing their interests to evolve unobserved. Through systematic evaluation on three Amazon datasets (Video Games, CDs & Vinyl, Movies & TV) using a strictly chronologically-aligned Gap-Synthesize Protocol, we reveal a consistent empirical pattern: sequence intent lifespan exhibits strict monotonic decay, culminating in a structural collapse when delta t exceeds one year, affecting all major SR paradigms equally. To address this, we propose DeltaGate, a lightweight, backbone-agnostic post-processing plugin featuring Gap-Driven Dimensional Gating. Operating on a frozen backbone, DeltaGate performs dimension-level representation routing: as delta t increases, it selectively suppresses volatile short-term dimensions while preserving stable long-term interests, falling back to a zero-initialized global prior. Counterfactual perturbation provides causal evidence that this routing is genuinely delta t-drive. On the critical >365d zone, DeltaGate recovers up to +106.1% Hit@10 using only 66K trainable parameters (<4% overhead). Systematic comparison with end-to-end retraining reveals two complementary adaptation regimes: full retraining maximizes accuracy but causes gate saturation and embedding drift. DeltaGate is the only configuration that maintains interpretable dimensional routing with 46x higher parameter efficiency and zero backbone drift.

List of all Short papers accepted for RecSys 2026 (in alphabetical order).

• RESA Causal Transformer Multi-Touch Attribution with Dual Debiasing and Explainable Visualization for Advertising Recommendation
  by Jiarong Zhang and Jing Gao

  Multi-touch attribution (MTA) aims to estimate the causal effect of advertising touchpoints on user conversions, providing essential support for budget allocation and advertising recommendation. Existing causal MTA methods often struggle to capture complex temporal dependencies in long touchpoint sequences and fail to adequately address confounding biases arising from both static user profiles and dynamic behaviors. To address these challenges, we propose CT-MTA, an end-to-end causal Transformer framework for multi-touch attribution. CT-MTA integrates a causally masked Transformer to model sequential dependencies, together with a Variational Autoencoder (VAE) and a Gradient Reversal Layer (GRL) to mitigate dual confounding biases. Furthermore, we design a counterfactual attribution module with a Mixture-of-Experts (MoE) architecture to generate personalized and interpretable attribution estimates. Experiments on the public Criteo datasets show that CT-MTA improves conversion prediction Area Under the ROC Curve (AUC) by 1.2% over the state-of-the-art method. In downstream budget allocation tasks, CT-MTA consistently reduces Cost Per Acquisition (CPA) and improves Conversion Rate (CVR) under limited budgets. Qualitative analysis further shows that CT-MTA alleviates over-attribution to historically favored touchpoints, leading to more efficient and fair budget allocation.

• RESAssessing Sentiment Semantics in KG-Based Explainable Recommender Systems
  by Alejandro Ariza-Casabona, Pol Pastells, Ludovico Boratto and Maria Salamo

  Sentiment information can enhance path-based explainable recommendation, yet advances are hindered by the lack of public benchmarks for sentiment-aware knowledge graph reasoning. We address this gap by systematically evaluating how different sentiment representation strategies and reward mechanisms affect utility, beyond-utility, and explanation quality across four real-world datasets under our open-source SAKG reasoning benchmark. Our exploration reveals: (1) sentiment-aware methods improve accuracy and diversity but reduce provider fairness, exposing a consistent trade-off; (2) under extreme sentiment imbalance, adaptive entropy-weighted meta-relations improve robustness in some domains, while naive sentiment sharing across heterogeneous relations causes negative transfer with considerable utility drop; (3) sentiment path consistency rewards enhance recommendation quality and explanation coherence, whereas terminal-only sentiment constraints improve fairness at modest utility cost. Our findings and open benchmark provide actionable guidance for designing sentiment-aware explainable recommenders and a foundation for future research in this underexplored space.

• RESBeyond Index-Only RoPE: Integrating Time and Order for Generative Recommendation
  by Xiaokai Wei, Jiajun Wu, Daiyao Yi, Reza Shirkavand and Michelle Gong

  In recommendation, however, position has two distinct meanings: an event’s order in the sequence and its wall-clock time. Existing approaches usually inject temporal information through auxiliary embeddings or relative attention biases, while vanilla RoPE captures only order. We revisit RoPE for generative recommendation and ask how to encode both time and order without losing the benefits of rotary attention. We present Time-and-Order RoPE (TO-RoPE), a simple family of designs that uses both signals in rotary embeddings with three lightweight variants: early fusion, split-by-dimension, and split-by-head. The key insight is that early fusion can cause interference between time and order inside the same rotary plane, whereas split allocation provides a cleaner inductive bias. Across public dataset and a proprietary industrial dataset, TO-RoPE consistently outperforms absolute, relative-bias, and single-source RoPE baselines.

• RESBookForYou: Leveraging BART-Generated Narrative Tropes for Content-Based Book Recommendations
  by Ritu Rajesh Kanchi, Ruhul Amin Hazarika and T. Gopalakrishnan Thirumoorthy

  This paper presents BookForYou, a hybrid content-based book recommendation system that integrates deep generative modeling with semantic retrieval. The system utilizes a fine-tuned BART model (139M parameters) to extract structured narrative trope tags from book descriptions, while Sentence-BERT (SBERT) performs initial retrieval from a catalog of 42,350 titles. Candidates are re-ranked using a late-fusion approach that combines semantic similarity with an Asymmetric Trope Overlap (ATO) metric. Evaluation on a held-out test set (n=89) demonstrates that the system achieves an NDCG@10 of 0.705, significantly outperforming SBERT-only (0.392, p < 0.001), BM25 (0.364), and popularity-based baselines (0.136). Ablation analysis identifies ATO as the dominant performance driver: trope-based re-ranking alone (0.691) improves upon SBERT by 76.3%, while the optimal hybrid configuration (alpha=0.55) provides an additional 2.1% gain (p=0.018). Furthermore, the system achieves a 92.7% trope match rate and a serendipity score of 0.679. By utilizing matched tropes as natural-language explanations, BookForYou offers a transparent and highly effective approach to narrative-driven recommendation.

• RESCandidate Retrieval for Provider Fairness under Severe Recommendation Imbalance
  by Patrik Dokoupil, Ludovico Boratto and Ladislav Peska

  Provider fairness in recommender systems is commonly addressed through reranking, which assumes that the candidate set already contains sufficient items from under-exposed provider groups. We study a setting in which this assumption breaks down: catalog representation does not translate into recommendation exposure, and minority items are largely missing from the candidates passed to the reranker. In this regime, methods such as FA*IR, calibration-based reranking, or minority boosting can only partially improve fairness because the items needed to rebalance the ranking are absent upstream. We address this bottleneck with FairSAE, an activation-steering method that shifts user representations toward minority items through a single contrastive direction in a latent space induced by a sparse autoencoder, before candidate retrieval. This pre-retrieval intervention surfaces minority items from the full catalog and can be combined with existing fairness-aware rerankers. Across the MovieLens-25M and Steam datasets, FairSAE-steered retrieval substantially improves the fairness-accuracy trade-off compared to standalone reranking. These results show that provider fairness under severe recommendation imbalance depends not only on reranking, but also on which items are retrieved in the first place. See https://bit.ly/fairsae for codes and data.

• RESChoosing What Matters: Query-Aware Multimodal Routing for Conversational Recommendation
  by Piao Huilin, Seoin Choi, Junbo Shim and Hayoung Oh

  Conversational recommender systems often benefit from evidence beyond dialogue context, and recent multimodal approaches have explored structured knowledge, textual semantics, and visual information to enrich user preference modeling. Building on MSCRS, which integrates multimodal graph representations with prompt learning, we propose QAMR-CRS (Query-Aware Multimodal Routing for Conversational Recommendation), a routing-based framework that explicitly estimates which modality-specific evidence is more relevant to the current query. QAMR-CRS forms a query representation from dialogue context and mentioned entities, uses it to route knowledge graph, co-occurrence, text similarity, and image similarity signals, and combines the routed evidence with a static multimodal prior before converting it into layer-wise prompt prefixes for the GPT-based recommender. We further apply entropy regularization to mitigate excessive dependence on a single modality. Experiments on ReDial and INSPIRED show that QAMR-CRS improves over MSCRS across major ranking metrics, with ReDial results averaged over three random seeds. Routing analysis further shows that entropy regularization mitigates single-modality collapse and promotes distributed routing.

• RESComparability in Recommender System Evaluation
  by Stefania Ionescu, Ilia Shilov and Florian Dörfler

  The increased awareness of the limitations of single-dimensional, user-centric evaluation motivated new emergent research in multi-metric and co-designed recommender systems. A key challenge here is integrating a multitude of potentially contradicting stakeholder perspectives and using them for meaningful comparisons between models. To this end, our work considers the full utility vector of stakeholders under alternative models, and uses social choice for aggregating individual benefits into a societal evaluation via social welfare functions. This allows us to bring entropy-based indexes used in algorithmic fairness that measures both group and individual unfairness as inequalities in benefits. Moreover, we use social choice theory to analyze comparability issues (i.e., the benefit variations of one stakeholder not being comparable with those of another) and suggest analytically informed solutions. We find that considering relative gains can restore comparability for fairness analysis while Nash social welfare provides a natural aggregation rule when social evaluation must remain invariant to stakeholder-specific rescaling. We also apply this approach to a real-world dataset and address limitations in practical implementations.

• RESCovariance-Aware Newton-Schulz Orthogonalization for Noise-Robust Sequential Recommendation
  by Jinxin Hu, Hao Deng, Haibo Xing and Lingyu Mu

  Autoregressive sequential recommendation trains transformer models on tokenized user behavior via cross-entropy. The Muon optimizer projects gradient matrices onto the Stiefel manifold through Newton-Schulz orthogonalization, ensuring uniform utilization of all parameter directions. However, we identify that this projection equalizes all gradient singular values to unity, collapsing the natural energy separation between signal and noise subspaces. In recommendation, where long-tail distributions, noisy implicit feedback, and exposure bias produce corrupted gradients, this equalization promotes noise directions to the same magnitude as signal directions—a directional amplification that magnitude-based techniques such as gradient clipping and SAM cannot address. We propose Covariance-Aware Newton-Schulz (CovNS), which estimates gradient directional stability via an EMA of the gradient outer product and attenuates volatile directions before orthogonalization. Experiments on a public benchmark and an industrial dataset show that CovNS outperforms Muon and all baselines across all retrieval metrics on both datasets with minimal overhead.

• RESDistribution-Level Contrastive Supervision for Generative Recommendation
  by Ziqi Xue, Dingxian Wang, Yimeng Bai, Shuai Zhu, Jialei Li, Xiaoyan Zhao, Frank Yang, Andrew Rabinovich, Yang Zhang and Pablo N. Mendes

  Recent generative recommenders improve scalability by retrieving items through token generation instead of traditional ranking over large candidate sets. Yet their training signals are still dominated by discrete code prediction, which overlooks the soft assignment information naturally produced by the tokenizer. This mismatch limits semantic transfer from the tokenizer to the recommender and may hurt overall optimization. We tackle this limitation by introducing a distribution-based supervision scheme for generative recommendation, where multi-level codebook probabilities are treated as soft semantic targets. On top of this design, we develop SODA, a plug-and-play alignment framework that adopts a BPR-style contrastive objective to align recommender representations with target-side distributional representations against negative ones. The proposed method enriches training with finer semantic cues while leaving the decoding stage unchanged. Experimental studies on multiple real-world benchmarks demonstrate that SODA consistently strengthens diverse generative recommendation architectures. Code will be available after acceptance.

• RESDual Conditional Diffusion for Generative Cross-Domain Recommendation via Disentangled Knowledge Transfer
  by Abhradeep Datta, Arin Gupta, Varun Tej Kasula and Ashok Singh Sairam

  Cross-domain recommendation (CDR) mitigates data sparsity by transferring knowledge across domains, and dual-target CDR further enables bidirectional transfer to improve both domains simultaneously. However, existing methods largely rely on linear augmentation or reconstruction, which may generate semantically inconsistent representations and noisy supervision. To address these limitations, we propose DCD-XRec, a Dual Conditional Diffusion framework that models cross-domain knowledge transfer as a structured and non-linear generative process. The model first learns expressive user embeddings with graph-based encoders and then decomposes them into shared and domain-specific components using learnable projection heads. On this basis, a bidirectional conditional diffusion module synthesizes target-domain embeddings conditioned on source-domain representations, enabling semantically coherent transfer. In addition, we introduce a dual diffusion objective with two complementary pathways: a real-to-real pathway for stable alignment and a real-to-augmented pathway for richer generative supervision. Experiments on multiple public benchmark datasets show that DCD-XRec consistently outperforms strong baselines, demonstrating the effectiveness of diffusion-based generative modeling for dual-target cross-domain recommendation.

• RESEnriching Graph-Based Modeling with Sequential Signals for Explainable Course Recommendation
  by Md Akib Zabed Khan, Dongsheng Luo and Agoritsa Polyzou

  Course recommendation systems (CRS) play a crucial role in guiding university students through their academic journey by assisting in course selection. At the same time, models that function as black boxes and offer little transparency in their decision-making process limit students’ trust and adoption of CRS. To address this, we propose Twiner, a TWofold sequential INtegration with knowledge graphs for Explainable next-basket Recommendation, that captures sequential patterns at two complementary levels: within the knowledge graph itself and through a recurrent model. This dual infusion enables richer contextual modeling for domains where the sequential nature of the data plays an important role. Our approach integrates graph attention networks (GAT) to model complex relationships (including sequential ones between items) and gated recurrent units (GRU) to capture sequential patterns in students’ course-taking behavior across semesters. To enhance explainability, we leverage the attention weights from the GAT module to provide path-reasoning-based justifications. We conduct experiments on four real-world datasets in the domains of course recommendation in higher education and online learning platforms, as well as grocery shopping. We show that our Twiner model performs similarly to or better than existing state-of-the-art models while also enabling richer explanations. Our findings highlight a transparent and effective solution where explainability does not come at the cost of performance.

• RESGradSup: Gradient Superposition for Personalised and Scalable LLM Recommendation
  by Kanishka Dandeniya, Chirath Dasanayaka, Daswin De Silva, Sam Saltis, Shalinka Jayatilleke, Nishan Mills and Harsha Moraliyage

  Large language models (LLMs) have demonstrated strong capabilities in recommendation tasks such as item, sequence, conversational recommendation, and explanation generation. However, LLM weights are typically shared across all users. Adapting these models to individual users remains a fundamental challenge that requires millions of trainable parameters, even when using finetuning methods such as Low-Rank Adaptation (LoRA). Building such personalised adapters would also require large volumes of storage and high computational overhead. To address this challenge of personalised and scalable recommendation, we propose the Gradient Superposition (GradSup) method. Built on the TinyLoRA architecture, GradSup is a closed-form method that computes per-user adapters in a single pass, achieving 15x speedup over iterative fine-tuning at matched parameter budget and 400x compression over full LoRA. Experiments conducted on movie, book and electronics benchmark datasets demonstrate that GradSup outperforms iterative fine-tuning to provide scalable and personalised recommendation that is consistently sustained above the frozen LLM backbone. Code is available at: https://anonymous.4open.science/r/GradSup-9860

• RESHierarchical GTV Estimation: Bridging the Gap between Container-Level Ranking and Item-Level Conversion
  by Zerong Lan, Fan Zhang, Chuang Chen, Tianchun Huang, Teng Zhang and Xingxing Wang

  Emerging e-commerce platforms, such as live-streaming and Online-to-Offline (O2O) services, inherently operate under a “Container-based” display paradigm. In this setting, a critical granularity mismatch exists: the platform performs ranking and estimation at the container level (e.g., livestream rooms or stores), whereas the actual transactions occur at the item level within these containers. This structural discrepancy complicates Gross Transaction Value (GTV) estimation, introducing unique challenges including high label variance, severe label sparsity, and dynamic item heterogeneity. To address these challenges, we propose a novel framework named Hierarchical GTV Estimation (HGE). HGE employs a Hybrid Set-Aware Encoder (HSAE) to model the dynamic composition of items and feature interactions. To mitigate label variance, we introduce a Proxy-Label Learning (PLL) strategy that decomposes post-click GTV into purchase probability, quantity, and unit price. Furthermore, a Hybrid Boosting & Bagging Strategy (HBBS) is designed to handle label sparsity by effectively combining item-level aggregation with container-level inference. Extensive experiments on large-scale industrial datasets demonstrate that HGE significantly outperforms state-of-the-art baselines, achieving a +0.63% lift in XAUC offline and a +1.42% increase in Revenue Per Search (RPS) in online A/B tests. HGE framework has been fully deployed into our main traffic.

• RESImpression Share Prediction: An Offline Evaluation Task for Ads Ranking Systems
  by Mohsen Malmir, Houssam Nassif, Danish Nasir Shaikh, Taher Rahgooy and Murat Bayir

  Offline evaluation is the main gateway before deploying ads ranking models to A/B testing in production. Standard offline metrics measure predictive accuracy, but are only a surrogate for advertiser value-the total conversions and revenue generated through the impressions advertisers receive. Advertiser value depends not only on prediction quality but on how impressions are distributed across objective buckets (advertiser campaigns organized by optimization goal, such as clicks, purchases, or video views). A model can show superior offline performance while shifting this distribution in ways that degrade advertiser value once deployed. No existing offline evaluation method provides visibility into these impression share shifts before deployment. We propose impression share prediction as an offline evaluation task: given a candidate ranking model, predict the distribution of impressions it would produce across objective buckets. The task is inherently counterfactual-the candidate has never served live traffic, pacing controllers have not adapted to it, and budget dynamics still encode the prior model’s equilibrium. We propose a structural causal model that captures how model predictions, advertiser budgets, and pacing jointly determine impression allocation, and show the counterfactual effect is identified from observational data. Building on this, we develop a statistical learning framework that predicts impression shares from offline model signals and current market state, trained on historical model deployments. On production data from multiple ranking model families, a Random Forest predictor reduces L1 prediction error by 49% over a constant baseline for models seen during training. For models held out from training we evaluate by time since the candidate first appeared in the live system; the first hour is the closest empirical analog to true pre-deployment, since the candidate has barely interacted with the system. In this regime the Random Forest falls below the constant baseline because the marketplace budget state still reflects the prior model’s equilibrium. An encoder-conditioned architecture that simulates a 2-hour rollout over recent marketplace dynamics recovers +22% L1 improvement in this regime.

• RESInteraction Modality and Trust: Investigating System-Driven Conversational Recommendation and Faceted Search
  by Laura Modre, Ahmadou Wagne, Thomas Elmar Kolb and Julia Neidhardt

  This work adopts an interdisciplinary approach to investigate how interactions with conversational recommender systems (CRSs) can foster user trust, compared to faceted search interfaces commonly used in e-commerce. We present an empirical comparison between a system-driven CRS and a functionally equivalent faceted search system, isolating the effect of interaction modality through a between-participants user study (N = 144). While perceived explainability and anthropomorphism were positively associated with trust across both conditions, no significant differences emerged in trust or its antecedents between conditions. In contrast, CRS users reported significantly lower perceived control. This difference reflects a design trade-off: the system-driven structure constrained user actions in order to ensure safe dialogue. Our results suggest that designers must balance the benefits of structured guidance against the agency cost of delegating control to the system.

• RESLLM-as-a-Judge for Evaluating System Responses in Conversational Music Recommendation
  by Seungheon Doh, Bruno Sguerra, Sergio Oramas, Elena V. Epure and Juhan Nam

  Conversational Recommendation Systems (CRS) aim to achieve two primary objectives: recommending relevant items and generating natural language recommendation responses, i.e., system utterances that present and justify suggested items to the user. While recommendation accuracy is effectively measured by established ranking metrics, the evaluation of response generation poses a more fundamental challenge. Although human evaluation remains the gold standard, its cost and scalability constraints have motivated the adoption of LLM-as-a-judge as a promising proxy, whose alignment with human judgment in the context of CRS remains an open question. In this paper, we present the first user study to empirically assess the reliability of LLM-as-a-judge for evaluating CRS responses. We sample 20 multi-turn music recommendation sessions and generate candidate system responses using four instruction-tuned LLMs (1B to 4B parameters), inducing variance in response quality across model scales. We collect 400 ratings from 20 domain-expert annotators, who evaluate each response across two orthogonal dimensions: Personalization Quality and Explanation Quality. Through bootstrapped correlation analysis with human preference scores, we demonstrate that LLM-based judges outperform all reference-based baselines. Furthermore, we analyze how judge performance varies according to model scale and conditioning information, providing practical guidance for deploying LLM-as-a-judge as a cost-effective proxy for human evaluation.

• RESPersonalized Recommendation Tool Learning via Autonomous Language Agents
  by Mingdai Yang, Zhiwei Liu, Weizhi Zhang, Yibo Wang, Hao Peng and Philip Yu

  Although large language models (LLMs) have recently gained traction in recommender systems due to their strong reasoning capabilities and extensive world knowledge, previous LLM-based agents suffer from hallucination and context-length limitations, and thus are not suitable for full-ranking recommendation tasks. To overcome these drawbacks, we propose an agent-based recommendation framework, Personalized Recommendation Tool learning via autonomous language Agents (PRTA), in which an LLM acts as a central planner interacting with multiple recommendation models as tools. The LLM-based agent is responsible for high-level reasoning and personalized tool selection, while traditional recommendation models perform full-ranking scoring, leveraging their scalability in modeling behavioral patterns. To support personalized tool selection, we design reflection mechanisms that enable the agent to evaluate and compare tools for each user based on user profiles and candidate ranked lists. Extensive experiments across three public datasets demonstrate the superiority of PRTA over traditional recommendation and LLM-based baselines in improving full-ranking recommendation performance. Our code implementation is available online.

• RESPosition Bias Induces Inconsistent Rankings in Listwise LLM-based Recommendation
  by Ethan Bito, Yongli Ren and Estrid He

  Large language models (LLMs) have emerged as promising listwise rerankers for recommender systems, but their reliability remains unclear. We show LLM-based rankings are not permutation-invariant, as reordering the same candidate set can change both the final ranked list and the pairwise preferences between items. We argue this behavior reflects a deeper structural problem, rather than simple output variability. Position bias perturbs local pairwise decisions, so the probability one item is ranked above another depends on input position. When aggregated across permutations, these local distortions can produce a pairwise preference system that is not consistent with any single global ranking. Instability in the final ranked lists is therefore a consequence of this underlying inconsistency. To study this phenomenon, we propose a multi-level evaluation framework that measures reliability at three levels, local instability in position-conditioned pairwise preferences, global inconsistency in aggregated pairwise preferences, and listwise instability across full rankings. We also consider a simple sequential inference strategy, Stochastic Greedy Selection, that reduces dependence on any single input ordering. Experiments across multiple models, datasets, and list lengths show mitigation methods do not improve reliability uniformly across levels. Methods that improve effectiveness or stabilize final rankings can still exhibit substantial inconsistency in the underlying preference structure. Our results show stability at the level of final rankings does not guarantee a coherent ranking process, and motivates evaluating on whether LLM-based recommenders define a consistent ranking function.

• RESPriCoRec: A Privacy-Aware Cloud–Device Collaborative Framework for Ad Recommendation under Feature Constraints
  by Dairui Liu, Zhongyi Lu, Jitao Lu, Aghiles Salah, Mete Sertkan, Roger Zhe Li, Changhong Jin, Barry Smyth, Xingsheng Guo and Ruihai Dong

  Privacy regulations increasingly restrict cloud processing of sensitive user data (e.g., age, gender), hindering traditional cloud-only recommendation models. To mitigate this challenge, we propose a Privacy-aware Collaborative cloud-device ads Recommendation framework (PriCoRec) that allows ad recommendations to be personalized while keeping sensitive data on-device. While separating recommendation into cloud-based and on-device stages enables privacy-aware deployment, naive splitting suffers from degraded shortlist quality and inefficient on-device inference due to limited private features. We therefore design a collaborative framework that explicitly compensates for this feature gap. Our approach comprises a cloud-based pre-ranking stage using cloud-accessible features, and an on-device ranking stage that locally incorporates highly personalized features. We introduce a diversity regularizer to pre-ranking to improve candidate quality. Moreover, to ensure controlled on-device power consumption and computational cost, we incorporate a cloud-guided training mechanism that enhances the performance of the device model while keeping the model lightweight. Empirical results demonstrate that the proposed framework maintains strong recommendation performance while keeping sensitive features on-device.

• RESProbing the LLMs for Generating User Profile of Session-based Social Recommendation
  by Mengying Lu, Hai-Tao Zheng, Lan Zhou, Hong-Gee Kim and Guangxin Wang

  Session-based social recommendation (SSR) aims to enhance session-based recommendation (SR) by incorporating social network information, making it particularly useful in cold-start scenarios. Existing SSR methods predominantly rely on graph neural networks (GNNs) to model social relationships, which often introduces computational inefficiencies and fails to fully leverage rich semantic and behavioral information. In this paper, we propose UPRec, a user profile-enhanced SSR framework that integrates large language models (LLMs) to refine user representations from three complementary perspectives: (1) text-based history embedding, which extracts semantic features from users’ historical interactions using pre-trained LLMs; (2) prompt-guided LLM embedding, which generates dynamic user interest representations through task-specific prompts; and (3) similarity-based social embedding, which captures behavioral similarities in user-item interactions and social relationships. By leveraging LLMs for user profiling and employing cosine similarity to preprocess social relationships, UPRec eliminates the need for complex GNN-based modeling while preserving efficiency and effectiveness in session-based recommendations. Extensive experiments and ablation studies on benchmark datasets demonstrate that UPRec outperforms state-of-the-art methods in both recommendation accuracy and computational efficiency. Our code is freely available.

• RESScaling LLM-Enhanced Linear Autoencoders to Industrial-Size Catalogs
  by Maxim Skurikhin, Kiryl Liakhnovich and Oleg Lashinin

  LLM-enhanced linear autoencoders (L3AE) incorporate semantic item representations from large language models into collaborative filtering and demonstrate significant gains on long-tail items. However, L3AE requires dense n × n matrices, which makes it impractical for large catalogs – exactly where semantic enrichment would help most. We propose scalable formulations that factor the semantic similarity matrix into a compact low-rank form and compute all required precision-matrix products without ever building any dense n × n matrix. Our method preserves the algebraic structure of L3AE while replacing its dense collaborative precision with SANSA’s sparse approximate inverse, enabling efficient scaling to industrial-size catalogs. Experiments on several public datasets show consistent gains over multiple baselines, and these gains tend to grow in colder, sparser catalogs – a pattern consistent with semantic enrichment helping most where collaborative signals are weakest. Code and configs are available at https://anonymous.4open.science/r/scal3ae

• RESStructure-Preserving Projection for Mitigating Modality Bias in LLM-Based Sequential Recommendation
  by Tzu-Wei Chiu, Song-Duo Ma, Hsin-Yu Lin and Pu-Jen Cheng

  Recent LLM-based recommenders integrate textual and collaborative signals by projecting collaborative embeddings into the embedding space of the LLM. However, this projection can introduce modality bias that distorts the underlying collaborative structure and limits the usefulness of projected embeddings. To address this issue, we propose a novel structure-preserving projection approach that maintains the relational geometry of collaborative embeddings through dedicated structure-preserving losses. Comprehensive experiments demonstrate that our approach consistently improves recommendation performance, providing a more reliable path for LLM-based recommendation.

• RESSupport Gap: Selecting Fixed-K Candidate Sets for Retained Personalized Headroom
  by Teresa Zhang

  Two-stage recommenders often compare fixed-size candidate sets before running an expensive reranker or online experiment. Recall at K and spread diagnostics describe exact-target coverage or set geometry, but they do not answer a fixed-budget bottleneck question: once retrieval has selected K items, how much state-contingent choice remains for a richer downstream ranker? We study retained personalized headroom, the gap between the value of fine-state-specific choices and the value of the best single coarse-state choice within the same candidate set. We introduce Support Gap, or SG, a candidate-local witness estimate of this quantity. Under a candidate-local uniform approximation condition, SG admits an explicit sufficient-condition error bound for retained headroom. Empirically, we use this theorem as an interpretive lens rather than as a certificate for the evaluated sets: witness audits show informative within-set rankings and reasonable central fidelity, but worst-case error remains large and the uniform condition is not verified. Across four offline constructions on public datasets, selecting candidate sets by SG lowers retained-headroom regret relative to Recall@K, RidgeProxy, and HardProxy. Pooled normalized regret is 0.2839 for SG, 0.3902 for Recall@K, 0.3327 for RidgeProxy, and 0.3311 for HardProxy.

• RESSurface Matching in Skill Recommenders: Incomplete Evaluation Hides Simplicity Bias
  by Warre Veys, Matthias De Lange, Jens-Joris Decorte, Chris Develder and Thomas Demeester

  Skill-extraction encoders are typically evaluated with ranking metrics scored against human-annotated positives. At the scale of ESCO (13,939 skills), annotation is inherently incomplete, and unannotated labels are treated as uniformly wrong irrespective of semantic proximity. We show that the nowadays popular contrastive bi-encoder setup is prone to exhibit simplicity bias, with a tendency for exploiting surface-level lexical overlap between input and skill label as a shortcut. It appears that annotated positives and such shortcuts often coincide, such that ranking metrics remain high even when other top predictions are potentially semantically nonsensical. To analyze this phenomenon at scale, we use an LLM-as-a-judge to flag nonsensical unannotated predictions, validated against human labels. We then study two training-time mitigations: label augmentation with low-overlap positives, and hard negatives targeting the shortcut which reduce judged nonsense while preserving ranking performance. Our findings show how incomplete annotations can silently degrade skill recommender quality, and give practical diagnostic and training-time tools for this setting, enabling us to release a more robust fine-tuned model to the community.

• RESTowards Efficient Hyperbolic Representation Learning for Recommender Systems
  by Tendai Mukande and Noel E. O Connor

  Real-world relational data in recommender systems (RSs), particularly in domains such as e-commerce, often exhibit hierarchical structures, such as linking User → Product Category → Sub-Category → Item. Message-passing models such as GNNs, which propagate information across nodes, have been applied in most existing RS models. However, repeated averaging in message-passing for complex hierarchical data often leads to oversmoothing and oversquashing, where node embeddings collapse to similar vectors, reducing their discriminative power and degrading model performance. To address these issues, we propose HyperRec, a hyperbolic representation model for heterogeneous recommendation, which preserves the hierarchical structure in the embedding space and mitigates feature collapse. Experimental results on three real-world datasets show that HyperRec achieves superior performance with competitive efficiency. The implementation code will be open-sourced.

• RESUncertainty-Aware Gated Context Fusion for Next POI Recommendation
  by Soyoung Jang and Jaekwang Kim

  Next Point-of-Interest (POI) recommendation requires modeling both sequential user behavior and rich contextual signals such as venue category and visit timing. While sequential models effectively capture visit patterns, they often underutilize these cues or rely on fixed-weight fusion that fails to account for varying informativeness across categories. We propose a modular gated fusion framework that integrates item, category, and temporal embeddings into any sequential backbone, with an uncertainty-aware gating mechanism that adaptively controls the contribution of contextual information based on category-level embedding variance. Experiments on Foursquare NYC/TKY and Yelp demonstrate consistent HR@K and NDCG@K improvements over strong baselines across multiple sequential recommendation architectures. We further analyze the learned gates to reveal when and how contextual signals contribute to next-POI prediction.

• RESUnderstanding ID-Text Complementarity in Sequential Recommendation
  by Liam Collins, Bhuvesh Kumar, Clark Mingxuan Ju, Tong Zhao, Donald Loveland, Leonardo Neves and Neil Shah

  Dense retrieval-based Sequential Recommendation (SR) systems increasingly leverage text features to represent items, and generally do so in one of two ways: (i) by completely replacing ID embeddings with text embeddings, or (ii) by carefully combining ID and text embeddings through complex fusion mechanisms. While these lines of work have produced strong recommendation performances, they offer conflicting perspectives on ID-text complementarity, or the extent to which ID and text embeddings specialize in modeling different signals: the former suggests a lack of complementarity, and the latter argues it exists but must be harnessed carefully. Moreover, neither workstream conducts an in-depth study of ID-text complementarity. We aim to clarify this picture by developing a rigorous understanding of the complementarity of ID- and text-based SR models in this work. Our study reveals that these models do learn complementary signals, meaning that either should provide performance gain when used properly alongside the other. Motivated by this, we introduce and evaluate a new, simple SR baseline that preserves ID-text complementarity through independent model training, then harnesses it via ensembling. Despite this method’s simplicity, we show it outperforms several competitive SR baselines, implying a third perspective on ID-text complementarity: both features are necessary to achieve state-of-the-art SR performance, but complex fusion strategies are not.

• RESWhen Fixed-Candidate Offline Evaluation Changes Model Selection in Two-Stage Recommenders
  by Teresa Zhang

  Two-stage recommenders use a first-stage generator to decide which items can be considered and a second-stage reranker to order the items that survive. Offline evaluation often removes this first-stage variation by scoring all systems on a shared candidate pool. That protocol is appropriate for measuring reranking quality on a fixed action space, but it changes the model-selection problem when compared pipelines differ in their generators, candidate budgets, or retrieval histories. We formalize the distinction between full-pipeline utility on a pipeline’s own candidates and reranking utility on an anchor-defined candidate pool, and we decompose the former into candidate survival and conditional ranking quality. In controlled registries built on ten public benchmark instances, fixed-candidate evaluation selects a different top pipeline from own-candidate evaluation in eight cases under the dataset-specific main anchor. On OTTO, Dressipi, and Diginetica, the top-1 disagreement persists under both tested K=20 anchors: OTTO and Dressipi show sign reversals, while Diginetica shows suppression of the own-candidate advantage large enough to change the selected winner. These results show that fixed-candidate scores are useful reranking diagnostics, but they should not be treated as full-pipeline model-selection scores when first-stage retrieval differs across systems.

List of all Past, Present, Future papers accepted for RecSys 2026 (in alphabetical order).

• PPFA Position Paper on Recommender Systems in the Era of Autonomous Agents
  by Aixin Sun

  For decades, recommender systems (RecSys) have been optimized to serve human users, leveraging behavioral data to predict preferences. However, the rapid deployment of autonomous agents powered by large language models (LLMs) introduces a paradigm shift: recommendation consumers are predicted to be increasingly a mixture of humans and authorized agents acting on their behalf, as a client-side proxy. This position paper reviews insights from prior human-centric RecSys research and outlines the transition to this more complex setting. We characterize the resulting tripartite interactions among humans, agents, and platforms, highlighting the dynamics that can arise across these relationships. This shift presents new opportunities for RecSys research, while introducing unique challenges for evaluation, alignment, and system design.

• PPFBackward and Inward: From Preferences to Personalization and the Safety Crisis Within Conversational Recommender Systems
  by Priyansh Singhal and Sumit Maheshwari

  Conversational recommender systems (CRS) have pursued progressively deeper personalization, dissolving the structural separation between system and user that characterized earlier paradigms by building rich models of individual preferences, memories, and emotional states through sustained dialogue. This paper argues that this trajectory carries a safety liability intrinsic to personalization itself: benign, organically accumulated user context degrades safety alignment across frontier LLMs through intent legitimation, memory-induced sycophancy, and cross-domain leakage, none of which require adversarial input. The vulnerability is compounded by a formal property of RLHF training that amplifies sycophantic tendencies, meaning the base LLMs on which CRS are built already carry a predisposition that personalization deepens. Emotionally vulnerable users are disproportionately affected, with population-level evidence of increasing emotional dependency and safety performance that degrades as user emotional intensity increases. This position paper traces the field’s evolution from static preference inference to conversational personalization, presents mechanistic and real-world evidence that the field’s core research objective and the safety vulnerability identified by the alignment community are structurally identical, and considers whether improved alignment, architectural decoupling, or scale can resolve the problem. We argue that the CRS community is well positioned to lead the development of personalization approaches that are safety-aware by design rather than safety-compromised by default.

• PPFBeyond Passive Preference Inference: User-Governed Preference Memory and Calibrated Initiative in Recommender Systems
  by Kaiwen Deng, Hao Jiang, Jiaxin Cheng and Wenming Yang

  Recommender systems enter their twentieth year still organized around passive preference inference: estimate latent interest from behavioral traces, then rank. Past: this abstraction was enormously productive—implicit feedback, matrix factorization, learning-to-rank, and large-scale offline evaluation defined the technical center of the field. Present: four mismatches now expose what prediction alone cannot settle—under-specified goals, evidence shifting from sparse clicks to natural-language preference statements, a wider action space than ranking, and evaluation that lags the proactive behavior it must capture. Future: the next decade should be organized around two first-class research objects, user-governed preference memory and calibrated initiative, composed inside a user-side / hybrid architecture and audited by an agency-centered evaluation that adds timing quality, memory correctness, and revocability to accuracy. To move from agenda to coordinated action, we ask the community to commit, within two RecSys cycles, to a shared reference schema, a mixed-initiative benchmark suite re-annotated from conversational-recommendation corpora, and a PPF-track reporting convention; as a first-mover deposit, we will publish a machine-readable JSON Schema for PreferenceMemoryRecord v0 as camera-ready supplementary material. These priorities do not replace ranking-centric work; they constitute the design surface on which the next decade of RecSys can be jointly read.

• PPFCodebook-Based Semantic IDs in Generative Recommendation: Enabling Interface, Emerging Bottleneck
  by Danil Gusak and Evgeny Frolov

  Codebook-based semantic IDs (SIDs), short discrete code sequences produced by quantization over item representations, made generative recommendation practical by turning catalog-scale retrieval into low-cardinality generation. Yet the same interface now concentrates the field’s hardest questions. We revisit the canonical two-stage SID lineage (RQ-VAE, R-KMeans, and close descendants) as a past-present-future story. Our thesis is that these identifiers succeeded as an enabling interface, but not yet as a stable theory of item identity for recommendation: one code must simultaneously preserve behaviorally useful similarity, protect identity under head-tail skew, survive catalog drift, align with language-model generators, and remain decodable under production constraints. We organize the present literature around five recurring fractures – objective mismatch, identity-versus-sharing tension, static codes in dynamic settings, LLM alignment tax, and serving feedback into identifier design – and argue that SID utility is regime-dependent: model capacity, catalog scale, and interaction density jointly determine whether semantic structure helps or is redundant. We close with a research agenda centered on recommendation-native tokenization, adaptive identity, explicit alignment protocols, and co-design of tokenization with serving.

• PPFEachMovie: archaeology of the first latent recommender system
  by John DeTreville

  This retrospective presents a preliminary archaeological reconstruction of EachMovie (1995–97), believed to be the first large latent recommender system; it became broadly available online a full decade before the Netflix Prize competition was announced. Running on two office PCs, EachMovie gave its users highly personalized movie recommendations by computing 20-dimensional latent vectors from its evolving 2.8M-vote dataset (97.63% sparse). EachMovie’s Joint Quartic Polak-Ribière iterative solver converged rapidly, and used a novel Hallucination-Free objective function to overcome the inherent problems of zero- or mean-filled matrices. EachMovie’s Early Voting data-augmentation strategy pre-loaded and stabilized the latent space every week by importing votes from professional reviews. EachMovie worked surprisingly well, cultivating a high degree of user engagement and trust, and noted movie critic Roger Ebert called it “uncannily accurate.”

• PPFForward and Outward: From Aggregate to Individual and Towards Human Understanding in Conversational Recommendation
  by Priyansh Singhal and Sumit Maheshwari

  Conversational recommender systems (CRS) have made substantial progress leveraging LLMs for preference elicitation and recommendation. Yet recent evidence suggests the field is approaching a structural ceiling: (1) recommendation accuracy degrades with deeper elicitation through a recall-specificity asymmetry, (2) the field’s most widely used metric shows near-zero correlation with self-reported user satisfaction, and (3) evaluation infrastructure suffers from reliability failures, benchmark biases, and invisibility to deceptive behaviors. This blue sky paper argues that moving past these limits requires a shift along two axes. Forward: from modeling only what users prefer toward understanding how they express preferences and why they hold them, engaging constructs such as linguistic hedging, conversational style, and motivational structure that current systems discard, and shifting from aggregate to individual-level evaluation. Outward: toward transdisciplinary engagement with psycholinguistics and pragmatics, which offer operationalizable tools for interpreting what users implicate beyond what they state, and with behavioral economics and cognitive science, which establish that preferences are constructed through interaction rather than retrieved from stable internal states. We propose the intersection of psycholinguistics and cognitive science as the most productive starting point for modeling users as whole persons rather than collections of attribute preferences.

• PPFFrom Side Information and Knowledge Graphs to Large Language Models: Two Decades of Knowledge Integration in Recommender Systems
  by Claudio Pomo, Diego Baquero Sanz, Liam Claude Morris, Ludovico Boratto, Fedelucio Narducci and Tommaso Di Noia

  Over the past two decades, Recommender Systems (RSs) have undergone successive transformations in how they encode and enable external knowledge: evolving from constraints and hand-crafted rules, to side information and feature matrices, to linked data and knowledge graphs, and, most recently, to Large Language Models (LLMs). This Past/Present/Future retrospective views this trajectory as a sequence of representational translations rather than a series of outright replacements. We analyze how these translations have transformed the functional capabilities of RSs, the recurring design patterns across technological eras, and the new risks that arise as knowledge integration becomes generative and conversational. By analyzing representative contributions from the RecSys literature, we identify several persistent regularities. External knowledge is repeatedly leveraged to mitigate sparsity and cold-start problems; human-interpretable structure remains fundamental for explanation, user control, and system governance; and hybrid architectures systematically reappear whenever no single representation simultaneously satisfies both scalability requirements and demands for traceability and auditability. We argue that LLMs should not be seen as replacements for structured knowledge representations, but rather as a new interaction and mediation layer that requires explicit grounding in verifiable, structured sources. We close by outlining a cautious research agenda for 2026–2030, centered on the development of hybrid, inspectable, and accountable RSs.

• PPFFrom User–Item Edges to Exposure Allocation: Matching Evidence to Platform-Level Recommender Claims
  by Teresa Zhang

  Recommender-systems research has progressively enlarged the decision objects it treats as central, moving from user-item response prediction to ranking, sequential policies, counterfactual exposure, fairness, multistakeholder utility, and ecosystem dynamics. As that enlargement has unfolded, a recurrent inferential problem has become increasingly difficult to ignore: evidence that is genuinely informative about a scorer, ranker, policy learner, causal estimator, or simulator is often read as though it already justified conclusions about how a platform ought to allocate scarce exposure over changing user, provider, catalog, and constraint state. We distinguish edge-sufficient claims, for which user-local scores, rankings, policies, or exposure effects remain adequate outer objects, from allocation-native claims, for which the relevant object is exposure allocation over changing platform state. On that basis, the paper offers a selective primitive-level map of the field, a diagnostic for identifying allocation-native claims, a compact motivating example showing why accurate immediate-response evidence alone need not identify the allocation implied by a broader platform objective, and a claim card for allocation-facing work. The purpose is thus to keep recommender claims proportionate to the evidence advanced in their support.

• PPFSome People Are Worth Recommending For: Twenty Years of Progress, Gaps, and Opportunities For and With Children at RecSys
  by Robin Ungruh and Maria Soledad Pera

  Recommender systems play a pivotal role in shaping children’s digital ecosystems. Despite substantial progress in recommender systems research over the past two decades, work that considers children as the main stakeholder, or even acknowledges them as a user group, remains in its infancy. Taking the 20th anniversary of the ACM RecSys conference as a milestone for reflection, we present a comprehensive review of 20 years of research published at RecSys and its co-located workshops, mapping the evolution of research related to children. Through this historical lens, we take inventory of the community’s contributions and identify unique challenges that emerge when designing, evaluating, and deploying recommenders for young people. Drawing on these insights, we outline critical gaps in the current landscape and propose a research agenda for advancing child-aware recommender system research.

• PPFThe Underrated Catalyst: Advancing Critiquing Towards Steerable and Next-Generation LLM Recommendations.
  by Huanyu Zhang, Xiaoxuan Shen, Baolin Yi and Yinao Xie

  As an interactive form of recommendation, critiquing demonstrates exceptional value in enhancing system steerability and transparency; nevertheless, its strategic significance has long been severely underrated. To address the pronounced chasm between current academic exploration and industrial practice, we thoroughly investigate the underlying causes of this divergence. Furthermore, we illustrate the critical role of this paradigm in optimizing algorithmic ecosystems, dismantling information cocoons, and establishing enduring user trust. In the era of generative recommendation driven by Large Language Models (LLMs), the utility of critiquing transcends mere explicit feedback, evolving into the foundational logic that facilitates the synergistic evolution of agents and humans at the architectural level. By integrating natural language interactions into the continuous self-reflection and evolution loops of agents, this mechanism necessitates a fundamental shift in recommendation logic, transitioning from unidirectional system delivery to human-machine collaborative decision-making. Looking forward, with the advent of the Agentic OS, critiquing is poised to elevate from a peripheral feedback tool to the core infrastructure that empowers users to steer complex intelligent systems. Consequently, by leveraging natural interactive interfaces and edge-based memory governance mechanisms, it ensures the long-term, precise evolution of user preferences while rigorously safeguarding privacy sovereignty. Ultimately, by reconstructing the theoretical framework surrounding the value of critiquing, we provide strategic guidelines for constructing transparent, steerable, and sustainable recommendation ecosystems.

• PPFTowards the Human-Centered Study of Recommender System Providers
  by Elizabeth McKinnie and Robin Burke

  Providers are essential stakeholders in recommender systems, but they have been under-studied in the recommender systems field. While prior work has examined topics of fairness and diversity in ways that represent provider-side concerns, so far the field has resisted conceptualizing providers as users and applying human-centered methods to understand and improve their experience. We identify five areas for future study required to bring providers into the fold as first-class users of recommender systems: impacts, interfaces, explanation, governance, and design. We propose concrete research questions for study, referencing existing work in RecSys and adjacent fields.

• PPFUser-Controlled Intent Layers for LLM-Mediated Personalization: A Research Agenda for Recommender Systems
  by Jiahao Liu, Mingzhe Han, Guanming Liu, Weihang Wang, Dongsheng Li, Hansu Gu, Peng Zhang, Tun Lu and Ning Gu

  Recommender systems have long been organized around a platform-centric architecture in which each platform independently constructs and optimizes its own internal representation of the user. As large language models (LLMs) increasingly become entry points to digital services, however, personalization may be reorganized around a different logic, one in which an LLM-based assistant mediates user intent and coordinates across multiple downstream platforms. In this position paper, we argue that the central transformation is a redistribution of representational control: from platform-internal profiling toward more user-addressable forms of intent mediation. We distinguish among platform profiles, LLM profiles, and a broader user-controlled intent layer through which users may not only be modeled, but also participate in shaping how they are represented and served across contexts. On this basis, we develop a five-part opportunity and challenge framework encompassing profile white-boxization and privacy, profile-mediated negotiation and intent alignment, cross-domain fusion and representation design, LLM-native advertising and trust, and profile ownership and governance. We then organize a research agenda for the recommender systems community along these same five dimensions. Our central claim is that, in an agent-mediated ecosystem, the next phase of personalization will depend not only on better prediction, but also on making recommendation more controllable, interpretable, portable, and trustworthy.

• PPFWho Are We Recommending To? Recommender Systems in the Agentic Web
  by Himan Abdollahpouri, Kyle Kretschman, Sai Ravindranath, Jackie Doremus and Mounia Lalmas

  For two decades, recommender systems have been designed under the assumption that a human directly consumes each recommendation: receiving, interpreting, and acting upon it. The emergence of AI agents powered by large language models challenges this assumption. In the emerging Agentic Web, autonomous agents increasingly act on behalf of users, e.g., browsing, comparing, negotiating, and executing transactions, raising a central question: who is the consumer of a recommendation? In this position paper, we argue that the recommendation paradigm is undergoing a bifurcation. In delegable contexts, such as routine purchases, travel, and constrained transactional tasks, the primary operational consumer of recommendations is shifting from the human to the agent, requiring new optimization objectives, interaction protocols, and evaluation criteria. In experiential contexts, such as entertainment, art, and other subjective or high-stakes choices, humans remain the final judge of relevance, though agents may assist through pre-filtering and curation. We introduce a delegation spectrum that characterizes recommendation contexts along factors such as preference specifiability, outcome verifiability, and decision stakes, and we outline a research agenda spanning agent preference modeling, dual-audience optimization, and the emerging agent attention economy. We further discuss the implications of this shift for the design and evaluation of recommender systems.

List of all reproducibility papers accepted for RecSys 2026 (in alphabetical order).

• REPRAttacking and Defending Multi-Agent Collaborative Filtering Systems Through Connectivity
  by Anjun Hu, Hanting Xie, Saranya Govindan, Jas Kandola and Kurt Cutajar

  Multi-agent collaborative filtering systems coordinate autonomous LLM-powered user and item agents through natural-language interaction to refine preferences and generate recommendations. These systems inherit vulnerabilities from both their data-driven nature and multi-agent interactions, which manifest in distinct ways. Understanding how connectivity modulates vulnerability in these systems could facilitate the development of more robust recommendation pipelines. In this work, we adapt attacks and defenses from the general multi-agent systems (MAS) literature to the agent-based CF setting, evaluating them under systematically varied connectivity in the AgentCF framework, where CF connectivity is characterized along two axes: (i) candidate count (the number of item candidates per turn per user, measuring user-side interaction density) and (ii) catalog concentration (the degree of item catalog overlap across users). Our contributions include: (1) Adaptation: we reproduce MAS-inspired attacks and defenses in the agentic CF domain, confirming partial transferability of original observations. (2) Characterization: we characterize how the two aspects of connectivity shape attack and defense outcomes, revealing role asymmetries between user and item agents, non-monotonic temporal dynamics in attack efficacy, and divergent patterns across dissemination and extraction attack goals. (3) Prediction: motivated by empirical findings, we assess the applicability of epidemic-inspired static metrics in ranking CF configurations by expected attack outcome without running full adversarial simulations, enabling cost-efficient robustness assessment. Implementation is available at https://anonymous.4open.science/r/connacf

• REPRAre We Really Making Progress in Group Recommendation? Unmasking the Tie-Breaking Illusion
  by Song-Duo Ma and Pu-Jen Cheng

  Recent group recommendation methods have reported strong improvements on standard benchmarks, but it remains unclear whether these gains always reflect genuine advances in modeling group preferences. In this paper, we show that several recent methods are affected by a systematic evaluation bias caused by the interaction between training-time score compression and evaluation-time deterministic tie-breaking. Specifically, an additional sigmoid transformation before the BPR objective can greatly increase tied top scores, making top-K metrics such as HR@K and NDCG@K highly sensitive to how ties are resolved. We revisit recent representative methods and their baselines onCAMRa2011 and Mafengwo under both group and user recommendation settings, and evaluate them with a tie-aware protocol that computes the exact expectation of HR@K and NDCG@K under uniform random tie-breaking. Our results show that many previously reported improvements shrink substantially under tie-aware evaluation, and the relative ranking of methods can change markedly. We further show that the additional sigmoid may act as implicit margin smoothing during optimization, and that temperature-scaled BPR can retain much of this benefit without inducing severe tie inflation. Overall, our findings highlight the importance of tie-aware evaluation for establishing reliable progress in group recommendation. The anonymized artifact package is available at https://anonymous.4open.science/r/RecSys-2026-Reproducibility-Artifact-A4CB/.

• REPRPersonalised Drug Recommender Systems based on Electronic Health Records: Reproducibility and External Validity
  by Archie Carpenter and Dietmar Jannach

  Recommender systems have achieved success in delivering personalised experiences across large-scale consumer platforms, particularly in domains such as entertainment and social media. Recently, there has been increasing interest in applying recommender system methodologies to domains with positive social impact, including healthcare. Although machine learning has demonstrated strong performance across a variety of clinical tasks, comparatively less attention has been given to recommender-style decision support for personalised care. The introduction and increasing adoption of Electronic Health Records (EHRs) has enabled the development of models that leverage patient histories to generate personalised clinical recommendations. In particular, EHR-based drug recommendation aims to predict safe and effective medication combinations using diagnoses, procedures and past visit information. In this work, we examine the current state of research in this area through an analysis and reproducibility study of representative models from the literature. We show that the field has grown rapidly in recent years and increasingly relies on complex deep learning architectures. However, our study reveals a stark gap between code availability and executability: despite nearly 88\% of papers providing public code, half of the executable pipelines failed entirely. We further demonstrate that only 2 out of 41 surveyed papers are fully reproducible, and we identify major questions regarding the real-world applicability of the proposed models. By highlighting these issues, we aim to provide a clearer understanding of the field and provide a foundation for more robust and clinically meaningful research in healthcare recommender systems.

• REPRWe’ve Already Been There: A Study of Data Leakage in Ephemeral Group Recommender Systems
  by Tom Lupicki

  We identify cross-interaction leakage, a data leakage problem in ephemeral group recommendation (EGR) evaluation that substantially inflates performance and obscures the task’s true difficulty. EGR models infer group preferences for ad hoc groups with no prior group-level history by learning from individual member interaction histories. Across three commonly used EGR benchmarks, we find that every group-item interaction is already present in every corresponding group member’s individual history, constituting up to 23.4% of all individual user-item interactions. Unlike previously studied data leakage problems in recommender evaluation, the held-out evaluation labels are directly present in the individual histories of the evaluated group’s own members. The task is effectively reduced to retrieving already-seen shared choices. We reproduce four EGR models on two benchmarks and evaluate them after removing all group-item interactions from individual user histories. Under this protocol, NDCG@20 drops by 18.3-42.9% on Weeplaces and 91.4-98.3% on Yelp across all models. This is not simply an artifact of data sparsity; randomly removing the same number of interactions per user yields drops of 4.7-18.8% and 34.5-63.7%, respectively. Further, all models are outperformed by a parameter-free heuristic under the standard leaky protocol. On a separately constructed dataset, we evaluate a user-inductive protocol to limit prior exposure to held-out group members’ individual histories, advancing toward leakage-free EGR evaluation.

List of all resource papers accepted for RecSys 2026 (in alphabetical order).

• RSCBinge Watch: Reproducible Multimodal Benchmarks Datasets for Large-Scale Movie Recommendation on MovieLens-10M and 20M
  by Giuseppe Spillo, Alessandro Petruzzelli, Cataldo Musto, Marco de Gemmis, Pasquale Lops and Giovanni Semeraro

  As Multimodal Recommender Systems gain interest, high-quality datasets with multimedia side information (text, images, audio, video) have become essential. However, much of the current literature relies on small-scale, undocumented, or non-public datasets. In this paper, we introduce M3L-10M and M3L-20M, two large-scale, fully documented and reproducible datasets that enrich MovieLens-10M and MovieLens-20M with multimodal features. Following a documented pipeline, we collect movie plots, posters, and trailers, extracting features via several state-of-the-art encoders. We publicly release raw data mappings, extracted features, and complete datasets to foster reproducibility and advance the field. Qualitative and quantitative analyses demonstrate our datasets’ utility across multiple perspectives. This work establishes a foundational resource for large-scale, multimodal movie recommendation. Our resource is available at: https://zenodo.org/records/18499145, with source code at https://github.com/giuspillo/M3L_10M_20M.

• RSChoploy: a Plugin-Based Inference Layer for Path-Based Explainable Recommendation over Knowledge Graphs
  by Ludovico Boratto, Gianni Fenu, Mirko Marras, Giacomo Medda and Alessio Sechi

  Path-based reasoning over knowledge graphs is a promising approach for explainable recommendation, as it justifies recommendations through entity-relation paths connecting user preferences to suggested items. However, existing implementations mainly target offline training and evaluation, while interactive deployment requires domain-specific handling of model loading, KG access, decoding, and explanation generation. We present hoploy, an open-source inference and explanation layer for the hopwise ecosystem that exposes pre-trained path-reasoning models through configurable APIs. hoploy supports stateless scenarios where users are unknown at training time and provide preferences only at request time. Its plugin architecture lets developers define request/response schemas, configuration files, and decoding logic that maps KG tokens into human-readable explanations. We release and demonstrate hoploy with POI and food plugins, assess their implementation effort and serving footprint, and provide documentation for extension. Resource: https://github.com/tail-unica/hoploy.

• RSCJTH: A Dataset for Evaluating Cold-Start and Temporal Dynamics in Job Recommendation
  by Julien Romero, Yann Millet and Éric Behar

  Online job‐matching platforms face a fundamentally different landscape from e-commerce or media recommendation. A job posting is typically live for only a few weeks, while an active candidate disappears as soon as they accept an offer. Besides, both sides may enter and leave the market multiple times over a career. These fleeting lifetimes (median 25 days for jobs, 10 days for candidates in our data) yield three compounding hurdles: (i) pervasive cold start for every newly posted vacancy and every first-time applicant, (ii) extreme interaction sparsity, and (iii) complex temporal overlap, because the window during which two entities can actually meet is narrow and highly variable (the median application process is of 8 days in our data).

• RSCscikit-rec: A Unified, Extensible Recommendation Library with scikit-rec-agent, a Conversational Interface
  by Shankar Sankararaman, Ivelin Angelov, Tin Nguyen, Jingyuan Zhang, Shivani Gowrishankar, Jaspreet Singh, Bowen Long, Qingbo Hu and Parvez Ahammad

  We present scikit-rec and scikit-rec-agent, two complementary open-source Python packages forming a composable, agentic recommendation framework. scikit-rec is a scikit-learn-style library built on a three-layer architecture (Recommender → Scorer → Estimator) that separates business logic, scoring strategy, and the underlying ML model, unifying five paradigms — ranking, contextual bandits, uplift modeling, sequential, and hierarchical sequential — under a single training and evaluation API. To our knowledge, it is the only library exposing IPS, SNIPS, DR, DM, and Replay-Match through a single evaluate() method shared across ranking, bandit, uplift, and sequential recommenders on the same dataset object (seven evaluators in total — six off-policy plus an on-policy baseline — and nine metrics), with implementation correctness verified by floating-point agreement against Open Bandit Pipeline on IPS, SNIPS, DR, and DM. Supported estimators span XGBoost, LightGBM, scikit-learn, and deep models (NCF, Two-Tower, NFM, DCN, SASRec, HRNN), with GPU optional. New paradigms plug in as a single Recommender subclass; we demonstrate this with Goal-Conditioned Supervised Learning (GCSL) for multi-objective recommendation, which compares directly against ranking, bandit, and uplift policies through the same evaluation API. scikit-rec-agent wraps the library as a conversational agent over eleven structured tools with a provider-agnostic LLM backend and deterministic URL-echo and import-scope checks on model output, turning plain-English data and goal descriptions into trained, evaluated pipelines without the user writing code. An internal version of the library has been in production at Intuit for 3+ years, powering 15 use cases across TurboTax and QuickBooks under sub-100 ms real-time latency contracts. Both packages are released under Apache 2.0 at https://github.com/intuit/scikit-rec and https://github.com/intuit/scikit-rec-agent, with reproducibility notebooks on MovieLens-1M and Amazon Books.

• RSCτ-Rec: A Verifiable Benchmark for Agentic Recommender Systems
  by Bharath Narasimhan and Karthik Narasimhan

  As recommender systems transition toward agentic, multi-turn conversational interfaces, evaluation paradigms have struggled to keep pace. Current benchmarks often rely on “LLM-as-a-judge” evaluations, which introduce subjectivity, high costs and inconsistency. We present ????-Rec, a benchmark for agentic recommender systems that replaces subjective evaluation with verifiable rewards and a reveal-tagged elicitation (RTE) mechanism that controls how task constraints surface during dialogue. By testing agents against structured catalog predicates and employing a pass^k reliability metric, τ-Rec provides a systematic test for consistent reasoning. Our evaluation of nine configurations across five model families — GPT-5.4, Claude Sonnet 4.6, Gemini 2.5 Flash, DeepSeek V4 Flash, Qwen3 32B and GPT-5 mini — reveals a steep reliability cliff, where even the best model achieves only ∼57% at pass^1 and ∼35% at pass^4, highlighting a critical gap in current conversational agent deployment. All code and data are publicly available at https://github.com/nbharaths/tau-rec.

• RSCWatchLens: A Configurable Platform for Online Video Recommendation Experiments
  by Deogyong Kim and Dongha Lee

  Studying how video recommender systems shape user behavior requires online experiments that link playback behavior with the recommendation conditions that produced it. Existing user-study infrastructure provides one or the other, but not both within a single experimentation workflow. We present WatchLens, an open-source platform that fills this gap. WatchLens adopts a modular architecture in which user interfaces, content sources, and recommendation policies are independently configurable, with policies assignable separately to the feed and the watch page, while a standardized logging layer attaches the recommendation policy and ranking position to every event at the time of recording. This design enables researchers to analyze how recommendation policies and ranking positions shape downstream playback behavior, session continuation, and navigation between the feed and the watch page, with the linkage between policy and outcome available within each event rather than reconstructed afterwards. We demonstrate WatchLens through a short-form video case study that holds the interface and content pool constant while varying only the watch-page policy, showing how the platform supports session-level comparison of recommendation effects on real viewing behavior. WatchLens is released as a publicly available, single-server deployable system for reproducible online video recommendation research.

• RSCWorkRB: A Community-Driven Evaluation Framework for AI in the Work Domain
  by Matthias De Lange, Warre Veys, Federico Retyk, Daniel Deniz, Warren Jouanneau, Mike Zhang, Aleksander Bielinski, Emma Jouffroy, Nicole Clobes, Nina Baranowska, David Graus, Marc Palyart, Rabih Zbib, Dimitra Gkatzia, Thomas Demeester, Tijl De Bie, Toine Bogers, Jens-Joris Decorte and Jeroen Van Hautte

  Today’s evolving labor markets rely increasingly on recommender systems for hiring, talent management, and workforce analytics, with natural language processing (NLP) capabilities at the core. Yet, research in this area remains highly fragmented. Studies employ divergent ontologies (ESCO, O*NET, national taxonomies), heterogeneous task formulations, and diverse model families, making cross-study comparison and reproducibility exceedingly difficult. General-purpose benchmarks lack coverage of work-specific tasks, and the inherent sensitivity of employment data further limits open evaluation. We present \textbf{WorkRB} (Work Research Benchmark), the first open-source, community-driven benchmark tailored to work-domain AI. WorkRB organizes 13~diverse tasks from 7~task groups as unified recommendation and NLP tasks, including job, skill recommendation, candidate recommendation, similar item recommendation, and skill extraction and normalization. WorkRB enables both monolingual and cross-lingual evaluation settings through dynamic loading of multilingual ontologies. Developed within a multi-stakeholder ecosystem of academia, industry, and public institutions, WorkRB has a modular design for seamless contributions and enables integration of proprietary tasks without disclosing sensitive data. WorkRB is available under the Apache 2.0 license at https://github.com/techwolf-ai/WorkRB.

List of all industry papers accepted for RecSys 2026 (in alphabetical order).

• INDA Framework for Value-Aligned Modeling and Robust Experimentation
  by Krishna Harsha Reddy Kothapalli, Angela Liu, Kamal Nayan Reddy Challa, Pavithra Seshadrivijayakrishnan, Yongpeng Yang and Mustafa Ispir

  When developing large-scale industrial recommender systems, a critical challenge is misalignment between offline model evaluation metrics and the final business objective. This disconnect often leads to inefficient development cycles where promising offline improvements fail to translate into online performance gains. This paper introduces a systematic framework to bridge this gap by applying novel value-aligned model evaluation metrics and reducing systematic errors in experiment setups at each decision-making stage. We focus on three experiment stages of the development process: offline experiments, low-traffic online A/B experiments, and high-traffic online A/B experiments. Our results from incorporating advertiser-defined value into a weighted offline objective improves Delivered Value by 4.7% on an O($B) surface and strengthens correlation between offline and low-traffic online results from a baseline of 0.1 (negligible) to now 0.65 (substantial) correlation; and our methods for reducing systematic biases in low-traffic A/B experimentation setups further improve low-traffic to high-traffic metric correlation from 0.53 to 0.93. Validated with data from over 100 large-scale offline and online experiments in both conversion (pConvs) and click (pCTR) prediction models, our framework provides a comprehensive guide for aligning machine learning development with core business objectives, leading to more accurate decision-making and significant (2x) improvements in engineering productivity.

• INDA Self-Triggered Agentic Push Recommendation System
  by Zhao-Yu Zhang, Qingying Chen, Chunyuan Zheng, Jing Zhou, Jian Sun, Siqi Chen, Leiying Chen, Chuan Zhou, Huiyou Jiang, Xin Tao, Haoxuan Li and Zhouchen Lin

  Push notification is a critical recommendation scenario on large-scale platforms, allowing the system to proactively reach users outside the application to improve long-term re-engagement. However, designing an optimal push system requires handling a complex action space for the “whether and when” delivery problem under strict system resource constraints, i.e., systems cannot naively evaluate every user at every second to find the optimal delivery moment. Existing solutions typically fall into two paradigms: the first is a two-stage approach that relies on offline user-level uplift modeling combined with integer programming solvers to allocate optimal delivery times and frequencies. The second periodically activates the push system to decide whether to send a push. However, the former cannot effectively utilize real-time information, and the latter results in resource inefficiency. In addition, such a multi-stage solution suffers from local optima. In this paper, we propose a proactive, self-triggered end-to-end agentic push recommendation system, which is already fully deployed at Douyin with over 1 billion users, allowing the system to generate push time and deternmine whether to send in a closed loop with both real-time effectiveness and efficiency. Specifically, the agentic system consists of two decision transformer-based agents: a planning agent that determines when to schedule the next push time with a gated ordinal regression method, and an action-execution agent that decides whether to send a push based on a trajectory reward. In addition, we further introduce a lightweight filtering agent to both control the computational overhead and act as a crucial safeguard against unreasonable planning behaviors. Extensive experiments demonstrate that this approach can both increase user active days by 0.3% and reduce push permission disablement by 1.9%. In addition, adding the filtering agent can reduce the computational overhead by 74%.

• INDA Unified Generative Re-ranking Framework with Adaptive Multi-objective Fusion
  by Heng Zhang, Yifan Gu, Wei Xu, Lei Cheng, Chuan Yuan and Hengrui Zhang

  In e-commerce recommendation, effective re-ranking must jointly optimize multiple business objectives such as clicks, add-to-carts, and purchases. Existing methods typically either employ multiple per-objective generators combined with static fusion through hand- crafted rules, or adopt non-autoregressive sequence generation with pre-defined objective fusion. Despite their differences, these approaches share a critical limitation: they treat multi-objective signals as statically fused, ignoring the fact that users have diverse and evolving intents in real time. To address this gap, we propose MAMGR, a unified generative re- ranking framework that enables end-to-end, sequence-level learn- ing of multiple objectives with dynamic weighting. MAMGR fea- tures a Multi-objective Adaptive Module (MAM) that learns person- alized objective weights in real time and shares them between the generator and evaluator, and a Dual-channel Context-aware Mod- ule (DCM) that jointly models local pairwise interactions and global sequence-level dependencies. Extensive offline and online experi- ments show that MAMGR consistently outperforms state-of-the-art methods, achieving +0.78% CTR and +1.43% GMV improvements while reducing CPU usage by 23.53% and incurring only +1.12% la- tency overhead. MAMGR has been deployed in Alipay’s production advertising system, serving hundreds of millions of users daily.

• INDAdvancing Relevance Measurement with Vision–Language Models for Web-Scale Search
  by Han Wang and Alex Whitworth

  Relevance evaluation plays a crucial role in personalized search systems to ensure that search results align with a user’s queries and intent. While human annotation is the traditional method for relevance evaluation, its high cost and long turnaround time limit its scalability. In this work, we present our approach at Pinterest Search to automate relevance evaluation for online A/B experiments using fine-tuned Vision-Language Models (VLMs). We rigorously validate the alignment between VLM-generated judgments and human annotations, demonstrating that VLMs can provide reliable relevance measurement for experiments while greatly improving the evaluation efficiency. Leveraging VLM-based labeling further unlocks the opportunities to expand the query set, optimize sampling design, and efficiently assess a wider range of search experiences at scale. This approach leads to higher-quality relevance metrics and significantly reduces the Minimum Detectable Effects (MDEs) in online experiment measurements.

• INDAlignment + Accuracy: The Cascade Reward Representation for Preranking
  by Hedi Xia, Dylan Zhou, Yali Bian, Yichu Zhou, Zili Li, Tianyou Wang, Bella Huang, Hongbo Deng, Piyush Maheshwari, Dafang He, Darren Reger, Bowen Deng and James Li

  Prerankers in large-scale recommender systems select candidates for a downstream ranker under strict latency constraints. In practice, teams combine accuracy metrics with alignment losses to train and evaluate prerankers, but what these quantities should target—and how to combine them—remains ad hoc. We derive the \emph{Cascade Reward Representation}: under mild assumptions on a fixed-retrieval, fixed-ranker pipeline, the expected change in user reward for preranker swaps with controlled overlap shift admits a calibrated first-order two-term representation $\mathbb{E}[\mathbf{R}^E – \mathbf{R}^0] = \alpha\,\mathbb{E}[\Delta\hat O] + \beta\,\mathbb{E}[\Delta N] + \mathcal{R}$, where $\Delta\hat O$ is a logged top-fraction overlap shift (\emph{alignment}: agreement with the main ranker’s selections), $\Delta N$ is a threshold-conditioned precision shift (\emph{accuracy}: engagement above a shared ranker threshold), and the remainder is bounded by $O(\mathbb{E}[\Delta^2]) + O_P(1/\sqrt{n})$. Both proxies are computable from production logs without running the main ranker on the full retrieval pool. This representation motivates a calibrated offline metric and a matching two-branch training loss for the model family studied in our production system. We validate the representation in a large-scale industrial recommender system. A calibrated linear combination of the two metrics raises experiment winner prediction from $45$–$50\%$ (accuracy-only) to $85\%$ and Pearson $r$ from ${\le}0.65$ to $0.84$ on held-out experiments. The matching training objective delivers $+1.43\%$ save engagement over an accuracy-only baseline and $+0.62\%$ over a heuristic alignment+accuracy production model in two-week A/B tests.

• INDAttending to the Core: Core-Task Attention for Recommendation
  by Jingyan Chen, Chenye Sun, Man Zhou, Yunhe Guo, Siyu Gu and Peng Jiang

  Industrial recommender systems typically support multiple busi- ness objectives through the integration of many specialized models. However, each individual model is usually optimized for a single primary target, such as conversions or purchases. For example, in advertising systems, models are often trained for OCPX-style objectives that tightly couple target prediction with bidding and revenue optimization. Since these target signals are often extremely sparse, multi-task learning (MTL) is widely adopted to leverage denser auxiliary tasks for additional supervision. However, exist- ing MTL approaches typically pursue balanced joint optimization across tasks, which may introduce task interference and degrade the performance of the core task. To address this limitation, we propose CoreAtt, a core-task-centric method that employs a novel attention mechanism to adaptively aggregate informative represen- tations from auxiliary tasks into the core task’s prediction tower. CoreAtt consists of two complementary attention pathways: (1) intra-sample attention, which models instance-level interactions among auxiliary tasks to produce context-aware fusion signals, and (2) inter-sample attention, which assesses each auxiliary sig- nal’s global reliability by comparing its prediction score against the population distribution. These two pathways are fused through a lightweight gating mechanism to enrich the representation for the core task. Notably, CoreAtt achieves strong performance even when built upon a simple Shared-Bottom architecture (CoreAtt-SB). We evaluate CoreAtt-SB on three public recommendation datasets, where it consistently outperforms strong MTL baselines while pre- serving auxiliary task performance. Moreover, online A/B tests on a leading short-video platform show that CoreAtt significantly enhanced platform revenue and advertiser value by 3.3% and 2.3%, respectively. Our code is available here1.

• INDBootstrapping Conversational Recommendation Agents At Spotify: Synthetic Data Generation and Self-Improvement Loops
  by Enrico Palumbo, Alexandre Tamborrino, Victor Ode, Ben Lacker, Adrià Casas Escoda, Jeremy Hopple, Marcus Better, James Leoni, Hugo Galväo, Hugues Bouchard, Mounia Lalmas, Jose Luis Redondo-García, Abenezer Abebe, Ann Clifton, Anton Blomberg, Henrik Lindström, Dani Doro and Christine Doig Cardet

  Conversational recommendation agents are emerging as a new paradigm for content discovery, enabling users to express complex intents through natural language (e.g., “recommend Italian indie artists I haven’t heard before” or “explain why they fit my taste”). A central challenge in building such agents is optimizing agent planning, i.e., deciding how to select, sequence, and invoke tools. This challenge is particularly acute in cold-start settings, where real user interactions are not yet available. We introduce a pipeline for multi-turn synthetic data generation and a self-improvement loop to address the lack of interaction data and the difficulty of optimizing agent planning in cold-start set- tings. The synthetic data pipeline transforms single-turn prompts into realistic multi-turn user–agent conversations, enabling sys- tematic evaluation of conversational capabilities before launch. The self-improvement loop then uses this data and evaluation feedback to combine variance-based contrastive optimization with iterative refinement through a coding agent, allowing the system to auto- matically identify and fix planning and tool-use errors. Our approach provides fine-grained insights into conversational capabilities, uncovers issues before deployment, and improves qual- ity by +8% on top of a highly optimized manual prompt, automat- ically resolving several planning and tool-use errors. The system has been productionized and significantly accelerated iteration cy- cles for the launch of a conversational recommendation agent at Spotify. Online A/B tests demonstrate its effectiveness, with +14% user listening, +5% increase in weekly active users, and a 5% reduc- tion in skip rate compared to a prior experience that only supports session refinement. Overall, this work provides a practical framework for acceler- ating the development of conversational recommendation agents in industry, addressing challenges that are becoming increasingly central to recommender systems and agentic applications as natural- language interfaces reach widespread adoption.

• INDBreaking the Loop: An Empirical Comparison of Strategies for Novelty and Freshness in YouTube Music
  by Srivaths Ranganathan, Zihuan Diao, Bernardo Cunha, Joshua L. Moore, Robin Dumas, Murat Goksedef, Yanwei Song, Mukai Lu, Gergo Varady and Tracy Pesin

  Continuously trained ranking models in music recommenders fall into feedback loops where previously consumed items dominate recommendations. This suppresses two distinct content classes: new releases (temporal freshness) and unlistened catalog items (novelty). Industry practitioners have a wide menu of interventions available, ranging from serving-time heuristics, training-data reweighting, architectural debiasing, to uncertainty-driven exploration, each of which are well understood in academic settings. But live systems offer challenges with continuously ingested content, interconnected components, and practical limitations that counteract the findings from academic research. We report results from off-policy online A/B tests for six interventions and a combination experiment across four conceptual layers (serving, training, architecture, exploration) on the YouTube Music homepage. All interventions modify the ranking model or the serving layer that consumes its scores; candidate generation and other upstream components are held fixed. We discuss key takeaways from our results: first, serving-time interventions on continuously trained systems are neutralized by the learning loop. Second, architectural debiasing reduces popularity dominance and improves diversity but does not create discovery, while carrying hidden integration costs. Finally, uncertainty-driven exploration interventions with a Spectral-normalized Neural Gaussian Process (SNGP) head produce the largest new-release lift, though they come with a measurable engagement or diversity tradeoff. We close with recommendations on which layer to intervene at, and the hidden costs of each choice.

• INDBuilding a User Foundation Model for the Open Web
  by Ivan Can Arisoy, Solal Vernier, Merwan Barlier and Blaž Škrlj

  User foundation models have demonstrated strong results in e-commerce and social recommendation, but most industrial deployments assume environments where user identity is stable and persistent. Open-web real-time bidding (RTB) operates on a structurally different data distribution: user identity is fragmented and non-persistent across browsing sessions, and the availability of browsing history depends on user privacy choices. Consequently, a significant portion of traffic carries no historical data, and available records often consist of relatively short, disjointed sessions. As a result, historical signals in this domain are typically represented as aggregated counters and recency buckets, leaving the sequential structure unexploited. To address this limitation, we present a user foundation model that applies self-supervised learning on user browsing histories and show that the learned representation improves multiple downstream production tasks, demonstrating the viability of this approach on the open web. We pre-train a Transformer encoder with masked language modeling and a sequence-level contrastive objective, then fine-tune it on the CTR prediction task. We optimize the encoder’s pre-training pipeline with an LLM-in-the-loop search over a curated catalog of reviewable, code-level edits (lifters), instantiating the LLM-as-optimizer paradigm in an industrial setting. The same encoder representation yields +1.197\% RIG on the production bid win-rate model and +1.354\% RIG on the production CTR ranker; a 12-day live A/B test confirms +1.29\% CTR, -0.89\% eCPC (80\% CI excluding zero on both metrics).

• INDCAPTS: Channel-Aware, Preference-Aligned Trigger Selection for Multi-Channel Item-to-Item Retrieval
  by Xiaoyou Zhou, Yuqi Liu, Zhao Liu, Xiao Lv, Bo Chen, Ruiming Tang, Guorui Zhou, Han Li and Kun Gai

  Large-scale industrial recommender systems adopt multi-channel retrieval for candidate generation, combining direct user-to-item (U2I) retrieval with two-hop user-to-item-to-item (U2I2I) pipelines. In U2I2I, the system selects a small set of historical interactions as triggers to seed item-to-item (I2I) retrieval across multiple channels. In production, triggers are often selected using rule-based policies or learned scorers and tuned channel by channel. However, these practices face two challenges: biased value attribution, which values triggers by on-trigger feedback rather than downstream retrieval utility, and uncoordinated routing, where channels independently select triggers under a shared quota, increasing cross-channel overlap. To address these challenges, we propose Channel-Aware, Preference-Aligned Trigger Selection (CAPTS), a framework that treats multi-channel trigger selection as a learnable routing problem. CAPTS introduces a Value Attribution Module (VAM) that credits each trigger with subsequent engagement from items retrieved through each I2I channel, and a Channel-Adaptive Trigger Routing (CATR) module that coordinates trigger-to-channel assignment. Offline experiments and large-scale online A/B tests on Kwai, Kuaishou’s international short-video platform, show that CAPTS consistently improves multi-channel recall offline and delivers +0.713% total app time spent and +0.586% average app time spent per device online.

• INDConAlign: Conditional Alignment Framework for Balancing Biased and Unbiased Recommendation
  by Jingcheng Zhang, Yihan Wang, Qi Song and Liyin Hong

  Industry recommender systems trained on observational data suffer from various biases that create filter bubbles, causing user interests to collapse into narrow categories and severely degrading long-term engagement. While utilizing unbiased uniform data for debiasing has shown promise, existing methods remain impractical for industrial deployment due to limitations such as neglect of factual(biased) recommendation performance and the substantial computational overhead. To overcome these limitations, we propose ConAlign (Conditional Alignment Framework), a conditional debiasing approach for industrial deployment. The key innovation of ConAlign lies in a discrete gating-based conditional alignment mechanism that selectively transfers knowledge from the biased tower to the unbiased tower. Following a selective intervention paradigm rather than universal correction, it seamlessly balances factual accuracy and unbiased preference estimation while supporting real-time streaming adaptation. To our knowledge, based on publicly available literature, ConAlign is the first architecture successfully deployed in a large-scale industrial system that utilizes a small fraction of unbiased random traffic for debiasing. Extensive offline experiments on three real-world datasets rigorously validate the effectiveness of our proposed framework, and the code is publicly available at \url{https://github.com/JcZhangzz/ConAlign}. Furthermore, large-scale online A/B testing on Kuaishou demonstrates significant improvements in long-term user engagement and interest diversity, with negligible latency overhead.

• INDDART – Drift Aware Ranking with forecasTing and uncertainty
  by Deep Nayak, Shreyas S and Sivaramakrishnan Kaveri

  Ranking systems in online platforms across domains must handle temporal drift where artifact distributions and contextual relevance evolve over time. Existing approaches address temporal drift using recency weighting, sliding windows, or forecasting techniques, however, they do not focus on integrating drift signals with ranking systems or handle the data variability introduced by the drift. We propose DART, a two-stage architecture combining a ranking model with a forecasting model that predicts artifact performance for a look-ahead window. An uncertainty-aware attention mechanism dynamically weights forecasted information based on prediction confidence, enabling robust adaptation without online updates to the ranker at serving time. Prior work uses uncertainty either to detect drift or to debias ranking; DART instead uses it as a control signal that modulates information flow from a forecaster to the ranker. Experiments on a large-scale e-commerce promotion ranking dataset show DART’s deployable offline configuration achieves +11.11% relative Mean Reciprocal Rank (MRR) during promotional events and +7.37% pre-event. On the public Coveo SIGIR dataset the same architecture delivers +22% MRR in both high and low-drift periods, showing dual-regime robustness across abrupt and gradual drift; the mechanism is task-agnostic and also improves a regression benchmark. Ablations demonstrate that attention-based integration of forecasted information substantially outperforms direct feature inclusion, reinforcing the importance of uncertainty-aware temporal adaptation in ranking systems.

• INDDRanker: A Transformer-Based Multi-Task Ranking Model for the Disney+ Homepage
  by Priya Nirmal Singh Khokher, Daniel Nemirovsky, Benjamin Whitesell, Ethan Sukrae Lee and Eric Wulf

  We present DRanker (Disney+ Ranker), a transformer-based multitask ranking model serving personalized recommendations on the Disney+ homepage. We describe the encoder-only architecture over enriched engagement-history sequences paired with a lightweight causal debiasing mechanism, and the inference stack that enables transformer-based ranking at scale. Through iterative development, we systematically evaluated more complex alternatives and found that simpler designs consistently outperformed them on offline metrics, qualitative behavioral evaluation, and in production. Deployed globally, DRanker drives over 1% lift in our key engagement metric and 0.5% lift in daily active users across all experiences powered by the ranker. We share practical lessons on data engineering, debiasing, and architectural choices that shaped our ranking system at scale.

• INDEGR: Embedding-Native Generative Retrieval with a Shared LLM
  by Xiaodong Liu, Congfei Zhang, Hsiang-Wei Chao, Siman Wang, Tong Zhao, Xiao Bai, Vincent Zhang, Jingxiao Ma, Zhe Liu, Wenfeng Zhuo, Zichu Li, Jitin Krishnan, Yunzhi Zhou, Yajun Wang, Jinchao Li and Yu Zhang

  Generative retrieval is increasingly popular in large-scale recommendation and advertising systems, yet current methods introduce practical complications. Semantic-ID methods rely on quantization, mutable identifier vocabularies, and token-to-item grounding; embedding-based pipelines train the item encoder separately from the query generator, which limits user-item alignment. We propose EGR, an Embedding-native Generative Retrieval framework for recommendation and advertising. EGR uses a single shared LLM to learn item representations from item metadata and user representations from interaction histories in one embedding space. Items are indexed directly as dense vectors, and user histories are encoded as dense retrieval queries. Joint contrastive training groups related items and aligns queries with their target items. We evaluate EGR on public benchmarks, industrial data, and live deployment. Besides outperforming published baselines on Amazon Reviews, on Snap DPA, EGR scales with data, handles cold-start items, and benefits from multimodal input. In production, EGR delivers a +3.12% conversion-rate lift, simplifying system design while improving retrieval quality and ad performance.

• INDEmbedding Subspace Partitioning for Dynamic Multi-Objective Retrieval
  by Shaobo Zhang, Alice Leung, Yunxiang Ren, Ping Liu, Yuchin Juan, Qianqi Shen, Benjamin Le, Jianqiang Shen, Chengming Jiang, Ko-Cheng Wang, Vidya Krishnamurthy, Caleb Johnson, Fedor Borisyuk, Luke Simon, Jingwei Wu and Wenjing Zhang

  Modern industrial recommender systems must optimize across competing objectives, balancing semantic relevance with business metrics such as engagement and revenue. While bi-encoders dominate large-scale retrieval due to their efficiency, they collapse these heterogeneous signals into a single static embedding space. This design creates a fundamental limitation: once trained, the retriever cannot adapt to shifting objective priorities at serving time without retraining. Moreover, joint optimization with multi-objective losses often induces interference between objectives, leading to suboptimal trade-offs. We propose \emph{Embedding Subspace Partitioning} (\emph{ESP}), a retrieval framework that decomposes the embedding into task-aware subspaces and replaces the single dot product with a weighted sum of per-subspace similarities, whose weights are tunable at serving time. For Transformer bi-encoders, ESP uses the model’s native end-of-sequence token as a segment delimiter, with segment-aware attention masking and position encoding resets to guarantee subspace isolation in a single forward pass. Serving is performed via GPU-accelerated exhaustive kNN over one concatenated index, eliminating the need for per-objective Approximate Nearest Neighbor (ANN) infrastructure required by multi-head approaches. We evaluate ESP on an open-source benchmark built from MS~MARCO~\cite{nguyen2016ms}. A single ESP model traces a broad Pareto frontier, consistently outperforming strong multi-task baselines across diverse operating points. In LinkedIn’s job matching platform (70M+ weekly users), ESP enabled dynamic retrieval reconfiguration and delivered significant key business metric lifts.

• INDEntity Representation Learning Through Onsite-Offsite Graph for Pinterest Ads
  by Zhimeng Pan, Jiayin Jin, Yang Tang, Jiarui Feng, Kungang Li, Chongyuan Xiang, Jiacheng Li, Runze Su, Chuizheng Meng, Siping Ji, Han Sun, Litian Tao, Ling Leng and Jamieson Kerns

  Offsite conversion data provide valuable signals for ads ranking because they reveal users’ commercial intent beyond their onsite activities. However, incorporating offsite data into large-scale ads models is challenging: conversion records are supplied by different advertisers and partner vendors, often contain sparse or inconsistent metadata, and exhibit behavioral distributions that differ from onsite ad engagements. In this work, we present a production-scale representation-learning framework for leveraging opt-in offsite conversion data in Pinterest Ads ranking models. We construct a heterogeneous onsite-offsite graph that connects various kinds of entities, such as users, ads, items, links, and advertisers, through onsite engagements and opt-in offsite conversions. We then apply TransRA, a revised variant of TransR, to learn scalable ID-based entity representations from this graph. TransRA designates one entity space as an anchor and maps non-anchor entity types into the anchor space, preserving relation-specific transformations while making the learned embeddings easier for different downstream ranking models to consume. We further study how to incorporate pretrained Knowledge Graph Embedding (KGE) representations into ads ranking models. Directly using pretrained KGE embeddings provides limited gains, suggesting a mismatch between graph pretraining objectives and ranking-model training. To address this challenge, we employ the Large ID Embedding Table technique and develop an attention-based KGE fine-tuning method that allows ranking models to adapt graph-derived entity embeddings jointly with the ranking objective while satisfying online serving constraints. We evaluate this framework in both the Ads Engagement Model for Click-Through Rate (CTR) prediction and the Ads Conversion Model for Checkout Conversion Rate (CVR) prediction. Offline experiments show consistent improvements in both models, and online A/B testing demonstrates gains in user engagement, conversion efficiency, advertiser value, CPA and CPC metrics. The framework has been deployed to full-scale traffic in both CTR and CVR models, showing that onsite-offsite graph representation learning can be effectively integrated into production ads ranking systems under industrial-scale constraints.

• INDFLUID: From Ephemeral IDs to Multimodal Semantic Codes for Billion-Scale Livestreaming Recommendation
  by Xinhang Yuan, Zexi Huang, Anjia Cao, Xudong Lu, Zikai Wang, Penghao Zhou, Chang Liu, Wentao Guo and Qinglei Wang

  Modern recommender systems rely heavily on ID-based collaborative filtering, where each item is represented by a unique ID embedding that accumulates collaborative signals from user interactions. Livestreaming recommendation, however, faces a unique challenge within this paradigm. A live room enters the candidate pool only while broadcasting, typically for tens of minutes, so its ID embedding never converges and ID-centric rankers fail to generalize. To address this, we present FLUID, the first framework to fully retire the item ID from a production livestreaming ranker. FLUID introduces a cross-domain multimodal encoder that is jointly trained on livestreams and short videos to produce discrete semantic codes (called LUCID) for content-based item characterization. Next, FLUID applies a staged warmup training scheme to adapt the ranker to LUCID. It first leverages the ranker backbone for a late fusion of cold, slice-level LUCID embedding and ID embedding and then replaces the ID embedding with warm, room-level LUCID embedding before the final online incremental training. Deployed on our online livestreaming ranker with a cross-platform combined user base of over one billion globally, FLUID delivers significant A/B gains of +0.55% Quality Watch Duration, +2.05% Cold-Start Room Views, and +0.05% Active Hours, demonstrating the superiority of its ID-free design at industrial scale.

• INDFrom Exploitation to Balance: Efficient Retargeting Recommendation via DPO-Guided Slot Allocation
  by Jiangwei Deng, Xiruo Shi, Lifang Deng, Dan Wang, Linke Zhao, Yujing Song and Xiaoyi Zeng

  In e-commerce recommender systems, items that users have previously interacted with are frequently re-recommended, a practice known as retargeting recommendation. While retargeted items offer significantly higher conversion efficiency by leveraging verified user interests, their inherent predictability induces a systemic bias in conventional recommendation pipelines. Specifically, retargeted items are prone to redundant retrieval across multiple recall channels and systematic score inflation by ranking models that over-rely on historical behavior signals. This unchecked dominance results in recommendation homogenization and user fatigue. We term this phenomenon the retargeting homogenization trap, which ultimately erodes long-term platform value. To address this problem, we propose RetargetRec, a novel industrial framework that formulates retargeting recommendation as an independent optimization problem. RetargetRec introduces a retargeting-aware recall module to implement explicit scale control and a Direct Preference Optimization guided by Item-level Reward (DPOIR) paradigm at the ranking stage. By training on item-level preference pairs derived from a composite reward, DPOIR enables optimization beyond immediate feedback signals without the computational overhead of conventional reinforcement learning. Additionally, a preference-based slot allocation mechanism then jointly governs the number, positions, and ordering of retargeted items. Extensive online A/B tests on the Lazada platform demonstrate that RetargetRec effectively suppresses overexposure while delivering significant gains in both transactions (+2.3\%) and GMV (+4.4\%).

• INDFrom M Passes to One: A MapReduce-Style Bootstrap for Pointwise Off-Policy Evaluation in Production Recommenders
  by Ghulam Ahmed Ansari and Cheng Ding

  We present Bootstrap IIPS@K, a pointwise off-policy estimator for production recommender rankers built around a MapReduce-style bootstrap factorization. The factorization constructs all bootstrap pseudo-samples through per-tile binary masks in a single read of the logged data, collapsing -pass variance estimation to one pass at billion-row scale. Bootstrap IIPS@K runs directly on production logs and removes the dedicated random-session logging branch that prior industrial estimators required. The factorization com- poses with any bounded per-session statistic, which lets structural sequence-dependence assumptions be compared empirically from the same single read; on production logs the pointwise prior stays bounded through slate depth = 60 while the cascade prior di- verges at = 10. A non-parametric quantile calibration of model scores keeps Bootstrap IIPS@K stable across multi-objective re- weights between ranker generations. Deployed on two LinkedIn Feed surfaces, Bootstrap IIPS@K reaches 70% offline–online par- ity (16/23 promotions) and Kendall’s = +0.29 (= 0.04) on the broader 26-pair rank-correlation pool; the Li et al. (WSDM 2011) baseline on the same workload shows near-random rank correlation with online lifts ( =−0.13).

• INDFrom Placements to Pages: Modeling Recommendation Module Interactions with Sequential Transformer Neural Bandits
  by Xingming Qu, Peijun Zou, Jing Shao, Xin Yin, Albert Hsiung, Kathy Hu, Nitish Dhinaharan, Dan Schonfeld, Kevin Liu, Shawn Zhou, Binbin Li, Adam Ilardi, Jesse Lute and Julie Cheng

  Modern recommender systems often populate multiple placements on a page using independent bandit or ranking models, implicitly assuming that placement decisions are independent. This assumption overlooks interactions among recommendation modules and can lead to suboptimal page-level outcomes. We reframe multi-placement recommendation as a sequential decision problem in which each module selection is conditioned on previously selected modules. We propose the Sequential Transformer Neural Bandit (STNB), a production framework that learns sequence-aware page representations for contextual bandit optimization. For latency-aware deployment, STNB uses a decoupled serving architecture with configurable autoregressive truncation, applying strict sequential inference to high-visibility placements and serving the remaining placements with a single batch scoring pass. STNB further incorporates a multi-objective bandit layer to align serving-time decisions with both engagement and downstream commerce outcomes. Offline evaluation shows improved predictive quality over an independent-placement neural contextual bandit baseline, and large-scale online A/B tests on eBay View Item surfaces demonstrate consistent gains in promoted-listings revenue while preserving broader marketplace outcomes. Ablation and latency analyses show that multi-objective optimization and truncated sequential serving are important for real-world deployment.

• INDGenerative Spatiotemporal Intent Sequence Recommendation via Implicit Reasoning in Amap
  by Sicong Wang, Ruiting Dong, Yue Liu, Bowen Zheng, Jun Meng, Jie Li, Shuaijun Guo, Yu Gu, Fanyi Di and Xin Li

  Real-world user behaviors are rarely isolated atomic actions but exhibit intent flows with spatiotemporal causal chains. To provide holistic service schemes, we focus on the task of Generative Spatiotemporal Intent Sequence Recommendation (GSISR), which aims to generate intent sequences that are both logically coherent and physically executable within complex spatiotemporal contexts. While Large Language Models (LLMs) offer strong reasoning potential for GSISR, their industrial deployment is hindered by significant inference latency and spatiotemporal hallucinations. To bridge this gap, we propose a generative framework GPlan that internalizes LLM reasoning into lightweight models through two key innovations. First, to achieve reasoning under strict latency constraints, we introduce Progressive Implicit CoT Distillation, which compresses explicit reasoning processes into special think tokens, allowing small models to inherit complex planning logic without generating long reasoning text. Second, to address the disconnect between general knowledge and real-world constraints, we design Spatiotemporal Counterfactual DPO. By aligning the model with counterfactual scenarios, we improve its responsiveness to spatiotemporal context and reduce context-mismatched plans. Offline experiments and online A/B testing demonstrate that our approach improves sequence coherence and context responsiveness. Our implementation and the anonymized GSISR dataset are available at https://github.com/alibaba/GPlan.

• INDGRIP: Generation and Reasoning for User Profile Completion
  by Riwei Lai, Yu Xia, Li Chen, Beibei Kong, Lei Cheng, Chengxiang Zhuo, Zang Li and Chenyun Yu

  User profiles, such as age and interest tags, form the backbone of modern recommender systems. However, in real-world scenarios, user profiles frequently encounter the problem of incomplete profile data, restricting the effectiveness of downstream recommendation tasks. Although large language models (LLMs) have shown remarkable potential in understanding user profiles, existing methods mainly focus on extracting explicit user information from external data and generating profile summaries. Such shallow reasoning patterns often fail to accurately infer missing user attributes under sparse data conditions. In this paper, we present GRIP (Generation and Reasoning from Incomplete Profiles), the first foundation model tailored for comprehensive profile inference under data sparsity for recommendation. GRIP introduces a unified three-stage training framework: (1) continual pre-training on large-scale structured user data for robust correlation modeling; (2) decomposed chain-of-thought reasoning with iterative self-distillation to facilitate in-depth profile inference; and (3) reinforcement learning with multi-dimensional rewards to jointly optimize factuality and reasoning coherence. Comprehensive experiments on realistic benchmarks along with online business deployments have demonstrated that GRIP significantly outperforms LLM-based methods in completing missing attributes and delivers substantial business gains in downstream recommendation tasks.

• INDGenPage: Towards End-to-End Generative Homepage Construction at Netflix
  by Lequn Wang, Jiangwei Pan and Linas Baltrunas

  We present GenPage, an end-to-end generative approach to Netflix homepage construction that replaces the traditional multi-stage recommender stack with a single transformer. GenPage represents the user and request context as a prompt, and autoregressively generates the entire structured, multi-row homepage as the response. We adapt the LLM training recipe: pretraining on positively engaged production pages, followed by post-training via weighted binary classification (WBC) or reinforcement learning (RL). For industry-scale deployment, we introduce techniques addressing cold start, model freshness, business-rule enforcement, and serving efficiency. In online A/B tests against a mature, highly optimized production homepage recommender, the WBC variant of GenPage delivered a +0.24% lift on the core engagement metric we use for launch decisions (p < 0.001), while reducing end-to-end serving latency by 20%. Offline experiments yield two findings worth highlighting: enriching the prompt yields a larger improvement than scaling model capacity in our current regime, and RL post-training increases homepage diversity even though diversity is not part of the objective.

• INDGuess Where You Go: Generative Next Point-of-Interest Recommendation in Amap
  by Penglong Zhai, Bowen Zheng, Jie Li, Yifang Yuan, Yue Liu, Sicong Wang, Mingyang Yin, Tingting Hu, Shuaijun Guo, Fanyi Di and Xin Li

  Generative retrieval enables recommender systems to retrieve items by generating compact item identifiers, but scaling it to industrial scenarios remains challenging due to redundant or colliding token assignments and insufficient integration of heterogeneous item signals. These challenges are particularly critical for next Point-of-Interest (POI) recommendation, where models must represent structured spatial entities, capture sequential mobility patterns, and produce predictions consistent with real user behavior. We propose Gwhere, an end-to-end industrial framework that integrates semantic identifier (SID) generation with LLM-based generative next POI recommendation. Gwhere first learns discriminative POI SIDs through a contrastive residual-quantization tokenizer that aligns textual, visual, spatial, and collaborative signals. Based on these SIDs, Gwhere adapts LLMs to mobility scenarios via continued pretraining on enriched spatio-temporal corpora, supervised fine-tuning, and Exposure-Aware Kahneman-Tversky Optimization (EAKTO), a reinforcement learning objective for behavioral preference alignment. Experiments on public datasets and Amap’s large-scale industrial dataset demonstrate the effectiveness of Gwhere. The system has been deployed in Amap’s homepage service under high-concurrency and low-latency constraints. Long-term online A/B tests show improvements of 5.83% in P-CTR and 6.20% in U-CTR over the production baseline. The implementation is publicly available at https://github.com/alibaba/SimCIT.

• INDHIPA-Net: Heterogeneous Intent Projection Alignment for Feedback Calibration in CVR Prediction
  by Mengqing Ye, Jianwei Zhai, Bobo Cheng, Fei Pan and Peng Jiang

  Lead advertising is a contact-oriented conversion scenario in industrial recommender systems, where the target action is to initiate follow-up communication rather than complete an immediate transaction. A key challenge in lead CVR prediction is that implicit drop-off behaviors are frequent and informative, but their feedback polarity is not directly observed. Treating such drop-off evidence as fixed negative feedback or ordinary noisy implicit feedback can distort user contact intent. We formulate lead CVR prediction as a semantic feedback calibration problem for unfinished contact behaviors. To address it, we propose HIPA-Net, a Heterogeneous Intent Projection Alignment Network. HIPA-Net learns candidate-aware representations for fine-grained positive, negative, and drop-off feedback behaviors, constructs explicit positive and negative intent anchors from reliable feedback, and calibrates implicit drop-off evidence through anchor-specific projection alignment. HIPA-Net has been deployed in a large-scale industrial lead advertising system. Offline experiments show that HIPA-Net outperforms representative behavior modeling, multi-behavior, feedback-aware, and denoising baselines. Online A/B tests further show +2.184% CTR, +5.426% RPM, and -11.521% user-side negative feedback rate, demonstrating the effectiveness of anchor-guided semantic calibration for contact-oriented recommendation.

• INDHeterogeneous Ranking in Industrial-Scale Recommender Systems: A Case Study
  by Di Bai, Jintao Liu, Zhenwei Tang, Peifan Wu, Nada Al-Thawr and Luoshu Wang

  Heterogeneous recommendation feeds present complex challenges that extend beyond those found in highly homogeneous environments (e.g., music-only or video-only closed-ecosystem platforms). In Google Discover, a unified feed integrates diverse content sourced from the decentralized open web, including web articles, long-form and short-form videos, user-generated content (UGC), and beyond. Different content types exhibit distinct feature densities and user interaction patterns. Building a unified ranking model that sustains high performance across such heterogeneity, while avoiding negative transfer or majority bias, remains a significant industrial challenge. This paper presents an end-to-end case study on the industrial-scale multi-task ranking of heterogeneous feeds, grounded in real-world deployment. We introduce HA-MoE, a heterogeneity-adaptive multi-gated mixture-of-experts architecture that incorporates explicit heterogeneity context into both gating networks and expert representations. This approach enables effective specialization without significantly increasing operational overhead. To support reliable deployment, we introduce LENS, a lightweight observability framework that provides interpretable diagnostics of expert specialization and tracks this functional heterogeneity across continuous retraining. We evaluate our method using Dual-Level AUC (DL-AUC), a heterogeneity-aware evaluation metric that combines global ranking performance with cross-segment ranking correctness. Offline evaluations on a large-scale industrial dataset demonstrate consistent improvements over baseline models. Furthermore, online A/B testing confirms gains in feed activity and exploration metrics. Together, offline and online results validate the effectiveness of our approach for managing heterogeneity in industrial-scale recommender systems.

• INDHybrid Retrieval and Multimodal Reranking for Item-to-Item Recommendation in a C2C Marketplace
  by Yusuke Shido, Ryo Watanabe, Yuta Ueno, Yuki Yada and Shinya Yaginuma

  Item-to-item recommendations on item detail pages drive a substantial share of session conversions on e-commerce services, including consumer-to-consumer (C2C) marketplaces, where listings are user-generated, short-lived, and diverse in both images and text. Visual vector-search recommendation, in which items are encoded by a pretrained vision model and retrieved by approximate nearest-neighbor (ANN) search using a query vector, has become a common building block for such surfaces. Mercari, Japan’s largest C2C marketplace, deployed a SigLIP-based instance of this approach in production with material online gains. However, visual similarity alone cannot capture buyer intent hinged on textual attributes (e.g., brand, model number, and condition), leaving items with weak user-uploaded images, common in a C2C marketplace, poorly covered. This study extends the visual-only setup into a multimodal, two-stage similar-item recommendation system. The system has two production components: (i) a hybrid retriever that ANN-searches over both image and Japanese text embeddings, and (ii) a multimodal neural reranker that consumes visual and textual item embeddings together with other item metadata. Both components were evaluated via large-scale online A/B tests. The hybrid retriever increased purchases from the similar-item recommendation by 20.4\%, and the reranker further improved them by 12.7\%. We share architecture, feature design, and lessons learned from deploying this multimodal similar-item recommendation system.

• INDHypothesis-Driven Shelf Generation for Personalised Recommendation
  by Aleksandr V. Petrov, Tarun Chillara, Matthew D. Moellman, Lucas de Haas, Yabai Song, Alina Susoykina, Melissa Crawford, Gabriel Negash, Erik Franco, Tasnim Rahman, Binal Jhaveri, Shubham Bansal, Hugues Bouchard, Roberto Mirizzi, Mounia Lalmas and Aloïs Gruson

  Modern recommendation interfaces organise content into shelves: themed rows such as “More of What You Like” or “New Releases for You.” In production systems, these shelves are typically defined through hand-crafted templates coupled with dedicated retrieval logic. While effective for broad recommendation intents, this approach does not scale to the long tail of individual taste. We present a content-hypothesis-driven shelf generation system for Spotify Home that replaces fixed templates with natural language hypotheses describing what a personalised shelf should contain. The system has three phases: generating shelf hypotheses from user profiles, retrieving catalogue items that fulfil them, and aligning the final shelf by selecting coherent items and revising its title and subtitle. This separation decouples shelf planning from catalogue fulfilment, enabling independent optimisation of both stages, constrained generative retrieval over catalogue entities, and distillation of frontier LLM behaviour into compact models. Our production pipeline combines hypothesis generation, generative retrieval, candidate selection and shelf alignment, offline LLM-as-a-judge evaluation, and precomputed serving. We describe the end-to-end architecture and evaluate it through offline ablations and a large-scale deployment on Spotify Home. Results show that hypothesis-driven shelves substantially expand personalised recommendation supply while achieving engagement comparable to the platform’s strongest curated shelves.

• INDIDProxy: CTR Prediction with Multimodal LLMs for Cold-Start Recommendation at Xiaohongshu
  by Yubin Zhang, Haiming Xu, Guillaume Salha-Galvan, Ruiyan Han, Feiyang Xiao, Yanhua Huang, Li Lin, Luo Yang and Yao Hu

  Content-driven platforms such as Xiaohongshu often leverage clickthrough rate (CTR) prediction models for recommendation. However, these models depend heavily on item ID embeddings, which perform poorly in item cold-start settings. In this paper, we present IDProxy, a production-scale system developed at Xiaohongshu to address this challenge. IDProxy leverages multimodal large language models (MLLMs) to generate proxy embeddings from rich content signals, enabling CTR prediction for new items in the absence of usage data. Through a lightweight coarse-to-fine mechanism, these proxies are aligned with the ID embedding space and trained endto-end with the ranking model, allowing seamless integration into production-facing pipelines. Extensive offline and online experiments demonstrate the effectiveness of the method, which has been deployed in 2025 in Xiaohongshu’s Content Feed and Display Ads features, reaching hundreds of millions of users daily.

• INDInterest Sequence for User Modeling in Industrial Short-Form Video Recommendation
  by Yuanzhen Lin, Diego Uribe, Yuan Shao, Ruixiao Sun, Yongle Cao, Zhimeng Jiang, Yuening Li, Yang Gu, Chuan He, Liang Liu and Sourabh Bansod

  Ultra-long user history modeling has been a highly effective approach in modern industrial recommendation systems, with most works heavily utilizing search-based methods and summarization based methods to map massive user interaction logs into latent user representation with algorithm designs to address the scaling challenge. Despite their success in user interest prediction, previous works are presented with significant scaling challenges due to the increasing computation cost in regards to sequence length and inherent information noise of item-based sequence. Furthermore, the lack of semantic interpretability and explicit interest segmentation often results in the loss of long-tail interest exploration. To address these limitations, we propose a novel Interest Sequence Modeling module and detail the real-world deployment within a production-scale ranking model serving billions of users. We introduce the User Interest Tree Profile (UITP), an explicit representation strategy that complements existing implicit chronological sequences by aggregating lifelong engagement metrics across a hierarchical taxonomy to extract explicit historical interest tokens of users. Through sequential modeling, this streaming asynchronous profile is processed into real-time interest-based sequence for model consumption. Offline evaluations demonstrate statistically significant improvements across all prediction targets and extensive live A/B experiments reveal core engagements gains, as well as highly beneficial ecosystem impacts: a $+2.43\%$ increase in exploratory views, a $+4.74\%$ boost in global inter-category diversity, and robust gains in fresh content engagement while maintaining low-latency serving constraints without large training overhead.

• INDLLM Reasoning for Subjective Tasks: Failure Modes, Mitigation, and Dynamic Reasoning Routing
  by Juncheng Dong, Ding Tong, Ishan Gupta and Yuyan Wang

  Recommendation systems fundamentally thrive on personalization, operating in a domain where “correctness” is rarely a binary truth but rather a matter of subjective human preference and sociocultural alignment. As Large Language Models (LLMs) are increasingly deployed as autonomous verifiers to evaluate complex safety and quality guidelines in these systems, they face a unique challenge: context-aware preference alignment. Recent advancements in Reinforcement Learning with Verifiable Rewards (RLVR) have significantly enhanced LLM reasoning capabilities, yet these gains are predominantly indexed on objective, mathematical tasks. In this paper, we investigate the generalization of explicit reasoning to subjective, human-centric industry rubrics using production datasets. We expose a fundamental vulnerability: forcing models to apply rigid, math-centric reasoning traces to subjective tasks actively degrades verification performance. Furthermore, applying standard RLVR to rectify this triggers a phenomenon we term reasoning collapse, where the reinforcement policy prematurely truncates reasoning trajectories in favor of rapid, heuristic guessing. To resolve this, we introduce a conditional length-penalized post-training algorithm. By mathematically intertwining strict verification accuracy with bounded reasoning length, we successfully stabilize the policy, halt reasoning collapse, and achieve state-of-the-art performance on production-scale verification tasks. Finally, we establish that the efficacy of a reasoning trace is deeply coupled with its socio-linguistic framing. We present preliminary synthesis results demonstrating massive performance variance across simulated demographic personas, and propose a novel mid-training architecture that dynamically routes reasoning through contextually aligned personas. Ultimately, this work provides a scalable algorithmic patch and a long-term architectural blueprint for aligning reasoning models with the friction of real-world subjective constraints.

• INDLLM-Based Re-Ranking for Real Estate Search
  by Nkateko Ntimane, Rafael Guedes, Tiago Cunha and Pedro Nogueira

  QuintoAndar Group operates the leading housing marketplace in Latin America for both rentals and sales. The platform replaces traditionally paper-heavy workflows with a fully digital experience, making housing transactions faster and more accessible to tenants, buyers, and landlords in the region. Finding the ideal home in such a vast catalog is inherently difficult. At the same time, the wide- spread adoption of conversational assistants is reshaping user ex- pectations: people increasingly want to express their needs through open, multi-turn dialog rather than rigid filter menus and faceted search. This shift is particularly pronounced in housing, where in- tent is multi-dimensional, context-dependent, and rarely reducible to a small set of structured constraints. To meet these expectations, we propose a Large Language Model (LLM) based re-ranker that augments a conversational recommendation system by reorder- ing retrieved candidates according to the nuanced, context-rich intent expressed across the user’s conversation. We additionally construct a large-scale offline evaluation dataset for conversational real-estate search, containing 960,000 query-item pairs constructed from both synthetic and production queries and annotated using an LLM-as-a-Judge framework with human validation. We validate our approach both offline, on this proprietary dataset, and online, through a production A/B test. Both evaluations show consistent improvements in ranking quality, including a statistically signif- icant increase in production of +5.3% in click-through rate and +4.8% in scheduled visits, demonstrating the value of integrating conversational context into housing recommendations.

• INDLLM-Based User Personas for Recommendations at Scale
  by Haoting Wang, Haokai Lu, Zheyun Feng, Jenny Huang, Yifat Amir, Gregory Hinkson, Ben Most, Zelong Zhao, Yixin Kelly Cui, Rein Zhang, Fabio Soldo, Yu Xia, Nihar Bhupalam, Minmin Chen, Konstantina Christakopoulou, Lichan Hong and Ed H. Chi

  Large Language Models (LLMs) offer unprecedented potential for enhancing recommendation systems through their world knowledge and reasoning capabilities. However, existing approaches often rely on structured IDs or offline processing, limiting semantic richness, real-time adaptability, and user-facing interpretability. In this paper, we introduce a novel framework that enables real-time generation of LLM-based user interest personas for a large-scale commercial video recommendation platform. Our method generates natural-language user interest personas that address the exploitation-exploration trade-off by combining the summarization of existing interests with novel topics, directly during serving. To overcome the computational challenges of online LLM inference at a billion-user scale, we design a cost-efficient architecture leveraging knowledge distillation, asynchronous inference, and input optimization via semantically clustered video representations. Extensive offline evaluations, user studies, and live A/B tests demonstrate significant improvements in viewer value. This work bridges the gap between high-level semantic understanding and industrial-scale recommendation, paving the way for more dynamic, explainable, and satisfying personalized experiences.

• INDLO-FAR: A Cost-Aware Local Filter for Sparse Feature Ranking in Industrial Ad Recommendation
  by Egemen Erbayat, Luis Duque, Sohini Roychowdhury, Mohammad Amin and Srihari Reddy

  Industrial ad recommendation models rely heavily on sparse, high-cardinality ID-list features that encode user histories and contextual identifiers. Each is backed by a dedicated embedding table, so these features dominate storage, training, and serving cost and must be revisited as traffic and downstream models evolve. Therefore, sparse feature ranking is not just an offline modeling problem but also a recurrent systems decision limited by compute budgets and iteration cadence. We present Localized Feature Ranking (LO-FAR), a CPU-only, model-agnostic workflow that ranks each candidate feature from its stand-alone held-out predictive signal using lightweight local estimators rather than the GPU-bound retraining loops of permutation- and stochastic-gate-based methods. On a production-grade dataset of more than one million logged interactions and 475 sparse ID-list features, LO-FAR completes ranking in approximately two CPU-hours and preserves downstream Normalized Entropy gains on CTR and CVR tasks that match or exceed shuffle-based importance, Binary Stochastic Neurons, and a coverage-based heuristic across budgets of 100–400 retained features. These budgets correspond to deterministic 40–75\% reductions in sparse storage. The contribution is a deployable workflow that shows how, under realistic cost and turnaround constraints, a simple local filter can be a stronger production choice than heavier interaction-aware alternatives.

• INDLearned Cross-Task Relationships in Multi-Task Models
  by Victor Zhang, Yiping Yuan, Florian Raudies, Bosun Adeoti, Brian Leung, Sanjay Surendranath Girija and Naijing Zhang

  We propose a framework that learns cross-task relationships in multi-task models by approximating the joint distribution of task labels through targeted pairwise relationships. This approach improves performance via transfer learning and enhances information extraction without the intractable complexity of modeling the full joint space. Although our framework applies to any multi-task system, we demonstrate its efficacy within YouTube’s production recommendation systems. Experiments across the Notifications, Homepage, and Watch Next surfaces show improvements in both accuracy and user satisfaction metrics. Finally, we propose a workflow template to facilitate broader future implementation.

• INDLong-term User Engagement Optimization through Model-agnostic Downstream Rewards Learning
  by Derek Wang, Filip Ryzner, Kelly He, Armando Ordorica, David Woo, Aditya Mantha, Liyao Lu, Usha Amrutha Nookala, Haoran Guo, Jiacong He, Olafur Gudmundsson, Matt Chun, Krystal Benitez, Dhruvil Deven Badani and Yijie Dylan Wang

  As recommender systems mature in the past few years, their optimization objectives have evolved from a primary focusing on short-term behavioral signals to a broader emphasis on long-term user engagement and retention. However, directly optimizing retention is difficult because return signals are sparse, delayed, and only partially attributable to earlier recommendations. Prior work has addressed this challenge with sequential modeling and reinforcement learning, but these approaches typically require task specific reward engineering, substantial computational overhead, and surface specific implementations that are difficult to generalize. In this paper, we present a unified, model-agnostic downstream reward framework for optimizing long-term user value in large-scale recommendation systems. First, we formulate the downstream reward learning problem and develop an offline screening framework to identify session level behaviors that are both observable early and predictive of future retention. We then propose several model-agnostic downstream rewards signals derived from observed user action patterns across multiple sources. We further discuss the engineering effort to productionize the proposed rewards derivations and challenges we faced when adding them to our ranking models. Online A/B experiments demonstrate consistent improvements in engagement and retention-related metrics, and the framework has been deployed across multiple Pinterest surfaces, including Homefeed, Related Pins, Search, and Notifications.

• INDM2R: End-to-End Implicit Scenario Discovery for Multi-Scenario and Multi-Crowd Ranking
  by Yuanhang Zhou, Chaoqun Hou, Wenzheng Fang, Jiaqi Zheng, Cheng Guo, Tong Liu and Bo Zheng

  E-commerce platforms typically host dozens of differentiated shopping scenarios, making multi-scenario modeling a de facto paradigm in industrial recommendation. However, classic methods rely on business rules to partition scenarios, which forcibly isolates similar samples across scenarios while failing to capture heterogeneous distributions within a scenario. Recent works either explore inter-scenario aggregation without addressing intra-scenario heterogeneity, or apply unsupervised clustering that frequently suffers from blurred cluster boundaries and prototype collapse due to the lack of supervision. In this paper, we present M2R, an implicit multi-scenario multi-crowd ranking framework: (1) it adaptively partitions samples into implicit scenarios driven by data distribution rather than business rules, addressing both cross-scenario collaboration and intra-scenario heterogeneity, and serves as a plug-and-play router that integrates seamlessly into mainstream MoE-based multi-scenario backbones; (2) it jointly optimizes the clustering router with the recommendation signal and two clustering-aware auxiliary losses end-to-end, mitigating the negative transfer caused by ambiguous cluster boundaries; (3) it introduces parameter-efficient LoRA experts that scale the expert pool to hundreds while preserving predictive accuracy and substantially reducing parameter counts. Online A/B tests on Taobao show that M2R lifts overall orders across 21 long-tail marketing scenarios by +6.3%, orders in the core Taobaomiaosha channel by +2.1%, and gross merchandise volume in Baiyibutie by +8.8%, demonstrating its practical value for industrial multi-scenario recommendation.

• INDMDGR: Masked Diffusion Generative Recommendation
  by Lingyu Mu, Hao Deng, Haibo Xing, Jinxin Hu, Yu Zhang, Xiaoyi Zeng and Jing Zhang

  Generative recommendation (GR) typically quantizes item embeddings into multi-level semantic IDs (SIDs) and generates the next item via autoregressive decoding. Despite competitive performance, this paradigm inherits three key limitations from language models: (1) autoregressive decoding struggles to capture global dependencies among multi-dimensional features at different SID positions; (2) a fixed decoding order implicitly assumes all users attend to item attributes identically; (3) sequential decoding is inefficient and struggles to meet real-time requirements. To tackle these challenges, we propose MDGR, a Masked Diffusion Generative Recommendation framework that reshapes the GR pipeline from three perspectives: codebook, offline training, and online inference. (1) We adopt a parallel codebook to provide a structural foundation for diffusion-based GR. (2) During training, we adaptively construct masking supervision signals along both the temporal and sample dimensions. (3) During inference, we develop a warm-up–based two-stage parallel decoding strategy for efficient generation of SIDs. Extensive experiments on multiple public and industrial-scale datasets show that MDGR outperforms 13 state-of-the-art baselines. Furthermore, by deploying MDGR on Alibaba Group’s online advertising platform, we achieve a 1.20% increase in revenue, demonstrating its practical value. The code will be released upon acceptance.

• INDMESH: Scaling Up Retrieval with Heterogeneous Content Unification
  by Jiaxing Qu, Yilin Chen, Junpeng Hou, Jinfeng Rao, Olafur Gudmundsson, Sai Xiao and Huizhong Duan

  Optimizing large-scale retrieval hinges on the ability to efficiently surface candidates across diverse content tiers. However, to capture segments such as fresh and long-tail content, modern systems typically resort to a fragmented “zoo” of specialized retrieval models. This operational complexity is attributed to a fundamental challenge in heterogeneous retrieval systems — the Scaling Bias of Heterogeneity — where model capacity gains do not apply equally across diverse content tiers. To bridge this gap, we propose MESH as a unified retrieval scaling framework that mitigates this bias through a modularized architecture integrated with gated bias correction. By partitioning the feature space into independent domains, MESH enforces a structural inductive bias that reduces interference between sparse-item signals and high-frequency engagement features. This protected gradient path leads to improved scaling behavior for sparse content, empirically validated by a 14× improvement in the power-law scaling exponent for fresh items. In online evaluations on Pinterest’s Related Pins platform — a billion-scale item-to-item recommendation system — these improvements translate into a +5.5% lift in fresh-item repins, alongside with 55% improvement in funnel efficiency and +0.46% improvement in user retention. Finally, our asynchronous serving strategy ensures production viability by delivering a 2.87× improvement in system throughput. Our findings suggest MESH as a promising paradigm for consolidating fragmented retrieval infrastructures into more scalable and ecosystem-aware backbones.

• INDMelo: A Production LLM-Powered Music Recommendation Agent
  by Shijia Wang, Da Guo, Qiang Xiao, Fanghui Bi, Weisheng Li, Dongjing Wang and Chuanjiang Luo

  We describe Melo, an LLM-powered music recommendation agent deployed on NetEase Cloud Music. Melo is structured as a deterministic five-node state graph over heterogeneous tools, with a prompt- and state-machine-driven orchestration policy rather than a fine-tuned controller. Two production failure modes drove the design: entity hallucination, where the agent commits to interpretations unsupported by the live catalog or user-behavior index, and long-tail degradation, where over-constrained requests collapse to generic popular fallbacks. We address them with two complementary mechanisms. Inference-time entity grounding repurposes the production search index as a verification primitive that gates entity decisions before they propagate downstream. Reflective retry verbalizes failure reasons from a broken tool chain and feeds them into the next planning step, so the system can relax or revise constraints rather than fall back blindly. A one-month online A/B test within NetEase Cloud Music’s playlist business unit reports an over 2 pp absolute lift in a primary playlist retention metric and a lift of over one minute in a core playlist engagement metric; offline ablation isolates a 7.8 pp absolute reduction in entity misidentification from the three-layer grounding stack on our evaluation set, and production triggered-session analysis shows reflective retry firing on 5.8% of sessions with 59% processlevel recovery (non-empty, error-free); constraint preservation is reported as a separate open evaluation question. The takeaway from this deployment is that, at production scale, progress on LLMpowered music recommendation hinges less on stronger language understanding than on runtime mechanisms that detect and recover from its mistakes.

• INDMend the Measurement Gap: Latent User Preference Modeling for Short-Form Video Recommendation
  by Shuo Chang, Jiangguo Zhang, Yueqi Wang, Zihuan Diao, Dapeng Hong, Ali Montazer, Joyneel Misra, Tomer Margolin, Sourabh Bansod and Ningren Han

  Recommender systems rely heavily on heterogeneous behavioral feedback to infer user preference. Although abundant, these signals are imperfect measurements: the same observed behavior can arise from different underlying states, such as genuine enjoyment, passive consumption, or inattention. The challenge is especially acute in short-form video, where watch-based signals are strongly affected by measurement confounders such as video duration — the same watch time can imply different levels of preference for videos of different lengths, while ratio-based metrics can systematically favor short videos. As a result, optimizing raw engagement can amplify measurement artifacts rather than improving user value. We propose a Factorized Latent Value Model (FLVM) for measuring user preference from heterogeneous behavioral feedback. The model treats observed behaviors as noisy measurements of a low-dimensional, factorized latent value state and uses structured output heads to model heterogeneous feedback signals. A restricted baseline path captures predictable variation from measurement-confounding features such as video duration, user propensity, and session context, while a routed latent path estimates preference-relevant value advantage. The resulting latent value score can be integrated into an existing recommender system as a ranking feature or ranking score. On YouTube Shorts, a major short-form video platform, this model improves offline metrics and lifts a primary viewer enjoyment metric by 2.67\% in online A/B tests.

• INDMoR: An Adaptive Retrieval Allocation Balancing Long-Term Interest and Short-Term Evidence
  by Yan Fu, Roni Peled, Brian Biesel, Werner Mostert, Raghuram Krishnaswami, Fei Wang and Qilin Qi

  Modern recommender systems often rely on multiple retrieval sources to generate candidate items, yet determining optimal quota allocation across these sources remains challenging. We present Mixture of Retrieval (MoR) bandit, a novel framework that dynamically optimizes retrieval allocation by balancing long-term user interests with short-term behavioral signals. Unlike traditional static approaches, our method employs a modified Thompson Sampling algorithm that combines a user’s affinity scores to third-party video channels with their recent behavioral signals, using exponential decay to prevent over-reliance on stale signals. The framework enhances personalization while promoting exploration of potentially valuable but under-exposed content channels. In a large-scale A/B test on Prime Video, MoR bandit achieved a significant 4% increase in third-party channel subscriptions and substantial improvements in user engagement metrics. At the time of writing this paper, MoR has been fully launched in production worldwide serving hundreds of millions of customers. Our approach provides a generalizable solution for multi-source retrieval optimization in large-scale recommender systems.

• INDMosaic: A Fleet of User Embedding Specialists for Recommendation at Meta
  by Zhiyuan Zheng, Xian Sun, Xiangyang Mou, Yujunrong Ma, Christina You, Michael He, Hrishikesh Paranjape, Aakarsha Agarwal and Hong Li

  User representation is one of the highest-leverage modeling problems in industrial recommendation systems: a single advancement in how users are encoded can propagate across retrieval, ranking, and integrity tasks at platform scale. Prior industrial user representation work builds either a single user model that emits one or more embedding vectors or a shared backbone with task-specific adaptation. In this paper, we present Mosaic, a foundational user modeling platform that employs a fleet of specialists to learn user embeddings. The fleet comprises four architecturally diverse model families – memorization-driven, dense-heavy, sequential-based, and CoTrain models – each focusing on a distinct facet of user behavior. We developed MRM (Multi-task Relations Mining) and CRL (Cosine Redundancy Loss) techniques to maximize the marginal information contribution of each new specialist. We also introduce CoEval and User Tower Zero-Out, new logging-free embedding evaluation framework that improves development velocity while preserving downstream-aligned accuracy. Our hybrid CPU/GPU, online-and-offline serving stack allows each specialist to choose the adequate serving strategy to meet the freshness, latency, and computational requirements. Mosaic delivers consistent and significant offline NE improvements in addition to online gains.

• INDMulti-Objective Ranking for Live-Streaming: Balancing Fresh and Delayed Signals with Segment-Aware Targeting
  by Xiaoyi Gu, Julia Tavares, Eder Santana, Carlos Mendoza-Cardenas, Nikita Mishra and Saad Ali

  One of the most challenging problems entertainment live-streaming services face in recommendation systems is that user behaviors are sparse and delayed, and interaction data exhibits bias for different user segments. Unlike e-commerce applications where user actions follow linear sequences, live-streaming viewers engage in multiple concurrent behaviors of watching, chatting, following, and spending, each occurring with varying delays. We address these challenges through three key contributions: 1) a delayed window approach that extends feedback collection beyond immediate responses, 2) a multi-model architecture that combines fresh and delayed signals, and a segment-aware targeting module that optimizes ranking scores differently across user lifecycle stages, and 3) Multi-gate Mixture-of-Experts (MMoE) integration that jointly models correlated targets while reducing model parameters by 41.9% compared to independent models. Online A/B testing demonstrates significant improvements, including a +0.09% increase in Daily Active Viewers (DAV), generating millions more annual active viewer days, and +0.56% increase in highly engaged viewers’ capped Average Revenue Per User (ARPU). Viewer-segment targeting achieved an additional +0.15% DAV improvement for newer and less engaged viewers, while MMoE enhancement added +0.08% overall DAV and +0.27% new follows. The experimental system processes ranking requests with low latency, providing a scalable approach for balancing multiple business objectives across diverse user populations. In addition, we tested the multi-model architecture on the mobile livefeed application and achieved a +1.12% increase in positive user-channel interactions (clicks, follows, and likes), demonstrating applicability beyond the primary use case.

• INDMulti-Probe Zero Collision Hash (MPZCH): Mitigating Embedding Collisions and Enhancing Model Freshness in Large-Scale Recommenders
  by Ziliang Zhao, Bi Xue, Emma Lin, Tianqi Lu, Mengjiao Zhou, Kaustubh Vartak, Shakhzod Ali-Zade, Tao Li, Bin Kuang, Rui Jian, Bin Wen, Dennis Van Der Staay, Yixin Bao, Xiujin Li, Chao Deng, Henry Wei, Songbin Liu, Qifan Wang and Kai Ren

  Embedding tables are critical components of large-scale recommendation systems, facilitating the efficient mapping of high-cardinality categorical features into dense vector representations. However, as the volume of unique IDs expands, traditional hash-based indexing methods suffer from collisions that degrade model performance and personalization quality. We present Multi-Probe Zero Colli- sion Hash (MPZCH), a novel indexing mechanism based on linear probing that effectively mitigates embedding collisions. With reasonable table sizing, it often eliminates these collisions entirely while maintaining production-scale efficiency. MPZCH utilizes auxiliary tensors and high-performance CUDA kernels to implement configurable probing and active eviction policies. By retiring obsolete IDs and resetting reassigned slots, MPZCH prevents the stale embedding inheritance typical of hash-based methods, ensuring new features learn effectively from scratch. Despite its collision-mitigation overhead, the system maintains training QPS and inference latency comparable to existing methods. Rigorous online experiments demonstrate that MPZCH achieves zero collisions for user embeddings and significantly improves item embedding freshness and quality. The solution has been released within the open-source TorchRec library for the broader community.

• INDMultilingual Semantic Retrieval for Apple Music Search
  by Vishalaksh Aggarwal, Kevin Sebastian, Vivek Kanojiya, Leo Le, Nick Tucey and Santosh Shankar

  Apple Music serves listeners across 150+ storefronts in dozens of languages, with a catalog that grows by 100,000+ new tracks daily. At this scale, search recall on misspelled, transliterated, and cross-lingual queries becomes a dominant driver of session quality, particularly for tail queries that account for the majority of unique queries. We present a multilingual semantic retrieval system built on a 305M-parameter Siamese bi-encoder fine-tuned from GTE-multilingual-base with curriculum-scheduled multi-objective training. The model is integrated into the search stack via a hybrid retrieval architecture that blends dense nearest-neighbor results with the existing token-based index using quantile distribution matching, enabling deployment without retraining downstream rankers. Offline, the model achieves a 69% relative improvement in Hit@10 over GTE-multilingual-base. In a worldwide online A/B test, the system delivers a 2.28% relative conversion-rate (CR) lift overall, an 86% reduction in the no-result rate, and gains across every storefront with no observed regressions. The improvement is concentrated where it is needed most: tail queries see a 7.93% relative CR lift, compared with 0.89% for mid-frequency queries and 0.14% for head queries—evidence that semantic retrieval improves recall on hard queries without disturbing well-served popular ones. To our knowledge, this is one of the largest search-quality improvements deployed on the platform.

• INDNextGen: A Multi-Objective Generative Re-ranking Framework for Taobao Recommendation
  by Longxiang Xiong, Chaoqun Hou, Zihao Zhu, Cheng Guo, Tong Liu and Bo Zheng

  Reranking is a critical stage in e-commerce recommendation that reorders items to optimize the final list as a whole. Existing genera- tive reranking methods suffer from two limitations: (1) they rely on next-item exposure prediction as supervision, neglecting page- level multi-objective signals such as total clicks or transactions; (2) autoregressive decoding incurs prohibitive latency, making full- candidate scoring infeasible under real-time constraints. We pro- pose Next-Gen, a generative reranking framework for page-aware multi-objective recommendation under strict latency budgets. Next- Gen introduces three key designs: (1) a full-candidate autoregressive generative framework via a global context-aware Encoder-Decoder architecture that jointly optimizes page-level multiple objectives while autoregressively selecting items across the entire candidate set, rigorously constraining inference latency for real-time deploy- ment; (2) a residual connection scheme for incremental reranking that feeds upstream ranking scores into both the item embedding layer and the candidate selection module, so the model only needs to predict nonlinear gains relative to ranking scores, improving con- vergence stability; (3) LLM-enhanced semantic embeddings from Taobao’s multimodal foundation model that enrich item represen- tations with domain knowledge at no additional inference cost. Online A/B tests on Taobao Miaosha show that Next-Gen achieves +18.85% GMV and +12.35% order volume over the strongest baseline. The system is deployed in production, serving hundreds of millions of users daily.

• INDPLAIN: An Explainable Generative Search System Enhanced by Multi-granularity Semantic Alignment
  by Guoliang Zhang, Weifan Wang, Junyao Zhao, Zhuo Li, Hongjing Zhang, Xiaobo Guo, Zhixin Zhai, Yonghui Zhao, Zhihao Wang, Jiayang Liu, Yingjie Cui, Jiwei Tan and Xuanping Li

  Industrial search platforms must efficiently retrieve relevant items from billions of candidates while satisfying both query relevance and user preferences. Generative Search (GS) has emerged as a transformative paradigm that reformulates traditional indexing and matching as an autoregressive generation task. However, most existing generative search models suffer from two critical deficiencies: (1) generated Semantic IDs (SIDs) often lack explicit semantic correspondences, undermining codebook interpretability; (2) unstructured codebooks impose a fully-connected search space, forcing the generator to navigate a highly entangled decoding path. To address these limitations, we propose PLAIN, which integrates Multi-stage Codebook Construction (MCC) and Unified Generative Retrieval (UGR). MCC leverages LLM-generated taxonomy and metadata labels, applying hard assignment for closed-set taxonomy levels and soft assignment for open-set metadata levels, transforming unstructured codebooks into interpretable hierarchical topic paths. UGR operationalizes the MCC schema by employing Symmetric Context Encoders (SCE) that align both query and item representations to the structured label space via knowledge distillation and semi-supervised hierarchical quantization, enabling consistent end-to-end generative retrieval. Extensive experiments and online A/B testing in Kuaishou’s live search system demonstrate significant improvements in user engagement and content consumption.

• INDPOEM: Partial-Order Enhanced Real-Time Sequential Modeling for Recommendation
  by Linxiao Che, Sun Yijia, Siyuan Lou, Shanshan Huang, Qiang Luo, Ruiming Tang, Han Li and Kun Gai

  Abstract Real-time recommendation systems face the challenge of dynamically changing user interests and contextual environments. Traditional sequential recommendation models rely on static historical click sequences, which struggle to capture real-time preference shifts and ignore the structured information embedded in the system’s internal ranking logic. This paper proposes POEM (Partial-Order Enhanced Modeling), a novel real-time sequential modeling framework that leverages the partial-order relations inherent in the recommendation cascade. POEM innovatively utilizes the real-time multi-task ranking scores (e.g., predicted CTR, watch time) from the preceding ranking stage as supervisory signals to construct dynamic partial-order sequences, thereby achieving fine-grained real-time interest modeling and aligning system objectives with user behavior. Our contributions are threefold: a partial-order guided sequence construction paradigm that augments traditional temporal sequences with a dynamically grouped and sampled sequence based on real-time ranking scores, enabling per-request interest reassessment; (2) a multi-objective score fusion mechanism that integrates various ranking signals through normalized rank-weighting into a unified quintuple representation; and (3) a hierarchical sample learning strategy combining system-preferred items (top-ranked) and user feedback (e.g., long-play videos) as positive samples, enhanced by graph-retrieved hard negative samples and a margin-based pairwise loss. Deployed in Kuaishou’s platform, POEM achieves significant online gains: +0.249% and +0.213% in average viewing time per user on the KS Single Page and KS Lite Page, respectively. Extensive ablation studies validate the effectiveness of each component, demonstrating POEM’s superiority in real-time responsiveness, recommendation accuracy, and content diversity.

• INDPROMISE: Process Reward Models for Unlocking Test-Time Scaling Laws in Generative Recommendations
  by Chengcheng Guo, Kuo Cai, Yu Zhou, Qiang Luo, Ruiming Tang, Han Li, Kun Gai and Guorui Zhou

  Generative Recommendation has emerged as a promising paradigm, reformulating recommendation as a sequence-to-sequence generation task over hierarchical Semantic IDs. However, current approaches face a severe challenge that we define as Semantic Drift, where errors in early, high-level tokens irreversibly divert the generation trajectory into irrelevant semantic subspaces. Inspired by Process Reward Models (PRMs) that enhance reasoning in Large Language Models, we propose Promise, a novel framework that integrates dense, step-by-step verification into generative models. Our framework utilizes a lightweight PRM to assess the quality of intermediate inference steps, and a PRM-guided beam search strategy that leverages dense feedback to dynamically prune erroneous branches. Most importantly, this method unlocks Test-Time Scaling Laws in recommender systems, demonstrating that by increasing inference compute, smaller models can match or surpass larger models. Extensive offline experiments and online A/B tests on a large-scale platform demonstrate that Promise, effectively mitigates Semantic Drift, significantly improving recommendation accuracy while enabling efficient deployment.

• INDPersonalizing Incremental Video Search with Hybrid Text and ID Embeddings
  by Vivek Kanojiya, Vishalaksh Aggarwal, Daeho Baek, Lyndon Kennedy and Xuetao Yin

  Incremental video search requires high-quality ranking after each keystroke, where intent is often underspecified (e.g., 1–3 character prefixes). We present a personalization system for Apple TV search that combines complementary semantic and collaborative signals at ranking time. Our approach learns two item embedding spaces: (i) a text-based multilingual encoder (TextEmb) fine-tuned on co-engagement triplets via contrastive learning, and (ii) an ID-based collaborative embedding model (IdEmb) trained on interaction-derived positives. At serving time, we construct user representations from recent watch history and inject text- and ID-based user–item cosine similarities into a pairwise XGBoost ranker. We evaluate the system with temporally held-out offline datasets and a three-week online controlled experiment. Offline, for sessions with user history, the personalized ranker improves NDCG@10 by 2.99% and MRR by 3.30% over the non-personalized baseline. Crucially, slice analyses show that personalization is most needed in incremental search, where intent is still forming: on ambiguous prefix queries (1–3 characters), NDCG@10 lift is +8.63%, versus only +1.46% on longer, more fully specified queries. Users with longer watch histories benefit more from personalization than newer users: NDCG lift rises from +2.13% for users with 1–5 history items to +4.37% for users with 51–100. This larger lift occurs even though baseline relevance is lower for long-history cohorts (NDCG@10 drops from 0.733 to 0.680), indicating that personalization adds the most value where default ranking underperforms. Online, treatment yields statistically significant gains of +1.14% tap-through rate and +1.23% conversion rate, with a 2.91% improvement in converted-item rank position. We further analyze coverage–precision trade-offs between semantic and collaborative embeddings through ablations isolating each signal, and evaluate embedding quality on a held-out corpus with LLM-judged similarity labels to reduce click/exposure bias.

• INDPinDCO: Whole-Page Aware Dynamic Creative Optimization at Scale
  by Yu Hao, Yuchun Li, Peimeng Sui, Meilin Liu, Tianyuan Cui, Hao Li, Zicong Zhou and Akanksha Baid

  Recent advances in generative AI have substantially accelerated the creation of high-quality ad creatives, dramatically expanding the number of candidate variants per campaign. This shift increases the need for scalable dynamic creative optimization (DCO) systems that can match creatives to the most relevant audiences under stringent latency and cost constraints. We present PinDCO, a production DCO system for ad creative retrieval and selection on Pinterest, a billion-scale visual discovery platform. PinDCO is built around a Creative Component Fusion Network (CCFN) that performs dynamic creative scoring by modeling each creative component (e.g., image, title, layout) with a dedicated tower, using component-specific hyperparameters to account for differing modeling complexity. The component representations are fused to predict a creative-level score conditioned on the ad-level pre- diction, and we improve training data quality via an exploration- exploitation strategy. To account for Pinterest’s waterfall grid layout, where a creative’s rendered size affects nearby content and session-level engagement, we introduce a Pixel-aware Adjustment Module(PAM) that adjusts scores based on creative size to encourage efficient screen real- estate utilization and better whole-page outcomes. To support the large volume of creative candidates, we further employ a light- weight pre-selection model for early pruning, and optimize serving efficiency through caching and dynamic batching. Extensive offline analyses and online A/B experiments demonstrate the effectiveness of PinDCO, yielding a +3.09% lift in ad Click-Through Rate(CTR) with positive whole-page metrics. With the strong performance, we launched PinDCO in the Pinterest Ads platform.

• INDProgressive Alignment of Recommender Foundation Model through Multi-Phase Post-Training
  by Oseong Choi, Hoeinn Kim, Jihoon Lee, Byungsoo Kang and Taeyeong Jang

  Foundation model(FM) for recommendation has shown strong ability to model long-horizon sequential user behavior. In practice, a single pretrained foundation model is often adapted to diverse downstream serving surfaces through Supervised Fine-Tuning(SFT). However, optimizing task-specific objectives such as clicks or likes does not necessarily align the serving policy with the business metrics that determine recommendation quality. We propose a three-phase progressive post-training framework that explicitly separates downstream adaptation from business-metric alignment. The adaptation stage is decomposed into Linear Probing(LP) and Full Fine-Tuning(FFT): LP first stabilizes randomly initialized downstream heads within a frozen pretrained representation space, and FFT then jointly specializes the full model for the target task. On top of this stabilized policy, Reinforcement Fine-Tuning(RFT) aligns the model with practical business objectives using a learned reward model. Rather than directly optimizing the serving policy on sparse business targets, we train the policy on dense implicit feedback and use business-metric supervision only for reward modeling. Offline experiments show that the progressive LP-FFT-RFT framework outperforms single-phase alternatives, and that reward-based alignment yields a stronger serving policy than directly using the reward model itself for ranking. Large-scale online A/B tests further show that the proposed framework improves production recommendation quality over a conventional non-foundation baseline. A reference implementation is available at https://github.com/webtoon/rec-fm-progressive-alignment.

• INDRECAP: Feedback-Driven Streaming Semantic User Profiles for Short-Video Recommendation
  by Ziyi Zhao, Xiaoyou Zhou, Xiao Lv, Yangyang Li, Chubo He, Zhao Liu, Jiayao Shen, Yuqi Liu, He Li, Chengyi Zhang, Jian Liang, Ming Li, Chongming Gao, Fuli Feng, Ruiming Tang and Han Li

  Language-based user profiles convert long behavioral histories into explicit semantic representations for recommendation. However, most profile generators are optimized in an open loop: they may summarize past behavior fluently, but are not directly trained to improve future recommendation. We study this problem in real-world short-video recommendation, where user behaviors continuously arrive as streams and profiles must be incrementally updated under limited capacity. This requires maintaining a consistent bounded profile state and constructing profile-targeted semantic feedback from industrial implicit behavior logs. We propose RECAP, an offline closed-loop framework for optimizing streaming structured semantic profiles with historical recommendation feedback. RECAP maintains each profile as a bounded structured memory by combining LLM-based semantic updates with deterministic lifecycle and capacity control. RECAP constructs profile-targeted semantic feedback by filtering label-consistent behavior pairs with an LLM judge and training a dual-tower evaluator whose matching score serves as a GRPO reward. Experiments on Kuaishou short-video data show that RECAP improves uAUC by 0.0084 and Recall@2000 by about 4.9% over the base generator. Further analyses confirm the benefits of feedback construction and policy optimization, and show more grounded refinement and user-level abstraction in profile updates.

• INDREEF: Real-time end-to-end Explore-Exploit framework for e-commerce Feeds
  by Devashish Gupta, Divay Jindal, Venkata Velugoti, Karthik Sundar, Vaishnav Chandak, Bhavuk Singhal, Vinit Rongata, Ravindra Yadav and Debdoot Mukherjee

  In modern e-commerce, recommendation feeds must balance exploitation (capitalizing on known user intent) with exploration (introducing novel catalogs to prevent “filter bubbles”). However, standard ranking architectures often suffer from negative transfer, where optimizing for discovery degrades conversion performance. We propose REEF (Real-time end-to-end Explore-Exploit Framework), an intent-aware, decoupled architecture that orchestrates traffic between specialized Explore and Exploit rankers. The two rankers are bound at the data-label level by the HandShake: an emergent synchronization mechanism, arising from conditional labeling strategy that operationalizes serendipitous discoveries into conversion pathways in real-time without added inference latency. Offline experiments demonstrate that REEF improves both diversity and relevance. Currently deployed at Meesho\footnote{www.meesho.com} for ~300 million monthly active users, REEF has significantly improved structural feed diversity and long-term business metrics.

• INDRankGraph-2: Lifecycle Co-Design for Billion-Node Graph Learning in Recommendation
  by Renzhi Wu, Zikun Cui, Junjie Yang, Tai Guo, Hong Li, Xian Chen, Li Yu, Ke Pan, Sri Reddy, Mahesh Srinivasan, Nipun Mathur, Haomin Yu and Hong Yan

  Graph-based retrieval at billion-node scale requires jointly solving three tightly coupled problems—graph construction, representation learning, and real-time serving—yet existing work addresses each in isolation. We present RankGraph-2, a framework deployed at Meta that co-designs all three lifecycle stages for similarity-based retrieval (U2U2I and U2I2I), where each stage’s requirements shape the others. Serving requires a co-learned cluster index to avoid expensive online KNN—this pushes index co-training into the training objective. Training benefits from the observation that similarity-based retrieval tolerates pre-computed neighborhoods, eliminating online graph infrastructure—this requires construction to produce self-contained data. Construction must also support hour-level refresh for item coverage. Acting on these cascading requirements, RankGraph-2 reduces hundreds of trillions of edges to hundreds of billions via subsampling with popularity bias correction, pre-computes multi-hop neighborhoods via personalized PageRank, and co-learns a residual-quantization cluster index that reduces serving computational cost by 83\%. This lifecycle co-design enables a simple architecture to achieve 3.8$\times$ higher recall than a GAT + Deep Graph Infomax model on a bipartite graph and 2.1$\times$ higher than PyTorch-BigGraph on item retrieval. RankGraph-2 delivers up to +0.96\% CTR and +2.75\% CVR, and has powered \textbf{20+ retrieval launches} across major surfaces.

• INDRecEvolve: A Knowledge-Driven Autonomous Agent System for Recommender Systems
  by Weidi Pan, He Ma, Shuhao Ye, Palaksh Rungta, David McPeek, Junyi Jiao, Arnab Bhadury, Mingyan Gao and Onkar Dalal

  The rise of agentic AI has catalyzed a shift toward self-iterating systems, opening new frontiers for the autonomous optimization of production recommender models. This paper presents the empirical validation of a knowledge-driven autonomous agent system, deployed directly on a production large-scale Two-Tower retrieval model. By delegating the entire research lifecycle, spanning idea generation, code implementation, offline training, and metric evaluation, to a continuous closed-loop autonomous framework, the agent system executed over 40 completed autonomous training runs from scratch. Executing these runs under rigorous production-scale evaluations, the system systematically navigated hidden architectural bottlenecks on the latest production model to achieve a breakthrough $\sim$20\% relative improvement in NDCG, a gain that translated directly to a +3.77\% increase in user satisfaction in live production traffic. Furthermore, the deployment exposed critical vulnerabilities in standard evaluation protocols, as the agent system autonomously discovered reward-hacking shortcuts. These findings prove that an autonomous pipeline can dramatically accelerate the pace of machine learning research and stress-test the rigorousness of underlying experimental infrastructure, while also exposing novel challenges such as reward hacking and redundant exploration of failed hypotheses.

• INDSMART: LLM-Augmented Hybrid Retrieval for Dynamic Product Ads
  by Congfei Zhang, Jingxiao Ma, Xiaodong Liu, Hsiang-Wei Chao, Siman Wang, Ge Liu, Shantanu Aggarwal, Vincent Zhang, Xiao Bai, Yunzhi Zhou, Yajun Wang, Zhe Liu, Jinchao Li, Yu Zhang, Rachel Liao and Meghana Missula

  Dynamic Product Ads (DPA) require retrieving relevant items from multi-million product catalogs, balancing two competing ob- jectives: retargeting (re-surfacing known interests) and prospecting (discovering new categories). While Large Language Models (LLMs) capture semantic intent better than traditional embedding models, deploying them at scale introduces prohibitive inference costs and lexical mismatch issues. Through controlled experiments on mil- lions of users, we demonstrate a critical retrieval decomposition: rule-generated queries excel at retargeting on a lexical BM25 in- dex, while LLM-generated queries excel at prospecting on a dense ANN index. Building on this, we propose SMART (SeMantic-aware Adaptive ReTrieval). To manage costs, a lightweight quality gate identifies coverage gaps in initial keyword results, adaptively rout- ing only the ∼10% of users who benefit from semantic prospecting to the LLM path. Offline evaluation demonstrates that this gated approach captures the bulk of semantic prospecting gains in Rel- evance Score while maintaining competitive re-targeting performance at a 90% reduction in LLM costs. Finally, in a 2-week online A/B test at Snap, SMART improved the ad conversion rate by +27.6% over a strong embedding-based baseline.

• INDSample Is Feature: Beyond Item-Level, Toward Sample-Level Tokens for Unified Large Recommender Models
  by Shuli Wang, Junwei Yin, Changhao Li, Senjie Kou, Chi Wang, Yinqiu Huang, Yinhua Zhu, Haitao Wang and Xingxing Wang

  Scaling industrial recommender models has followed two parallel paradigms: \textbf{sample information scaling}—enriching the information content of each training sample through deeper and longer behavior sequences—and \textbf{model capacity scaling}—unifying sequence modeling and feature interaction within a single Transformer backbone. However, these two paradigms still face two structural limitations. Firstly, sample information scaling methods encode only a subset of each historical interaction into the sequence token, leaving the majority of the original sample context unexploited and precluding the modeling of sample-level, time-varying features. Secondly, model capacity scaling methods are inherently constrained by the structural heterogeneity between sequential and non-sequential features, preventing the model from fully realizing its representational capacity. To address these issues, we propose \textbf{SIF} (\emph{Sample Is Feature}), which encodes each historical Raw Sample directly into the sequence token—maximally preserving sample information while simultaneously resolving the heterogeneity between sequential and non-sequential features. SIF consists of two key components. The \textbf{Sample Tokenizer} quantizes each historical Raw Sample into a Token Sample via hierarchical group-adaptive quantization (HGAQ), enabling full sample-level context to be incorporated into the sequence efficiently. The \textbf{SIF-Mixer} then performs deep feature interaction over the homogeneous sample representations via token-level and sample-level mixing, fully unleashing the model’s representational capacity. Extensive experiments on a large-scale industrial dataset validate SIF’s effectiveness, and we have successfully deployed SIF on the Meituan food delivery platform.

• INDSelf-Evolving Recommendation System: End-To-End Autonomous Model Optimization With LLM Agents
  by Haochen Wang, Yi Wu, Daryl Chang, Li Wei and Lukasz Heldt

  Optimizing large-scale machine learning systems, such as recommendation models for global video platforms, requires navigating a massive hyperparameter search space and, more critically, designing sophisticated optimizers, architectures, and reward functions to capture nuanced user behaviors. Achieving substantial improvements in these areas is a non-trivial task, traditionally relying on extensive manual iterations to test new hypotheses. We propose a self-evolving system that leverages Large Language Models (LLMs), specifically those from Google’s Gemini family, to autonomously generate, train, and deploy high-performing, complex model changes within an end-to-end automated workflow. The self-evolving system consists of an Offline Agent (Fast Loop) that performs high-throughput hypothesis generation to optimize for proxy metrics, and an Online Agent (Slow Loop) that validates candidates against delayed north star business metrics in live production. Our agents act as specialized Machine Learning Engineers (MLEs): they exhibit deep reasoning capabilities, discovering novel improvements in optimization algorithms and model architecture, and formulating innovative reward functions that target long-term user engagement. The effectiveness of this approach is demonstrated through several successful production launches at YouTube, confirming that autonomous, LLM-driven evolution can surpass traditional engineering workflows in both development velocity and model performance.

• INDSequenceO1: End-to-End Ultra-Long (100K) Sequence Modeling in Recommendation with Low-Rank Caching
  by Lin Guan, Jia-Qi Yang, Zhishan Zhao, Jiaqi Huang, Hangyu Wang, Longbin Li, Beichuan Zhang, Haonan Jiang, Jinan Ni, Xiangyu Fan, Xiaowen Li, Ziyao Ren, Yuhang Qi, Xiaolong Zhu, Xuanyuan Luo, Qiwei Chen, Yi Cheng and Lele Yu

  Modern short-video recommenders must exploit ultra-long user histories—often on the order of $10^5$ interactions per user—but are constrained by strict latency and training-throughput budgets. As a result, production systems typically truncate histories or rely on two-stage retrieve-then-rank pipelines, sacrificing long-term signals and breaking end-to-end optimization. While Stacked Target-to-History Cross Attention (STCA) enables end-to-end modeling up to the 10K regime, directly scaling it to 100K remains prohibitive for serving: its dominant target-conditioned cross attention is query-dependent and its cost still grows linearly with the raw history length. We present \textbf{SequenceO1}, an end-to-end framework deployed at full traffic on Douyin that scales long-sequence ranking to the \textbf{100K} regime at billion scale. At the model level, we propose \textbf{Sketch Attention (SA)}, which compresses an ultra-long history into a fixed-size, user-only sketch using learnable prototypes and \emph{prototype-wise} normalization (each token distributes mass over prototypes). We then perform target-conditioned reasoning at two time scales: STCA over a recent 10K suffix for recency and STCA over the fixed-size sketch for ultra-long signals, followed by lightweight fusion. At the system level, the user-only sketch enables cache-first reuse: it can be computed once and reused across multiple targets and consecutive requests, removing $n$-dependent computation from the ranking critical path on cache hits. We further improve efficiency with multi-request user-level batching in training and a fused FlashSA kernel for sketching under ragged batching. Together, these model and system optimizations make end-to-end 100K sequence modeling practical in production.

• INDShape Your Feed: An LLM-based Agentic System for Conversational Recommendation
  by Ziyun Xu, Bosen Ding, Ji Qi, Qingyuan Song, Jizhou Huang, Liwei Wang, Yue Zhang, Qichao Que, Yue Weng, Zhenheng Yang, Junfeng Pan, Jeffrey Santelli and Linhong Zhu

  Industrial recommendation systems predominantly adopt a passive ranking paradigm that infers user preferences from implicit behavioral signals (e.g., clicks, dwell time) rather than explicit, natural language inputs. As a result, users experience a persistent discrepancy between their explicit interests and what passive behavioral algorithms deliver, limiting their ability to express nuanced preferences or steer their feed in real time. To address this growing gap between how recommendations are optimized and how users wish to articulate their interests, we present Shape Your Feed (SYF), an LLM-based agentic recommendation framework that enables real-time, multimodal co-curation of content. SYF employs a three-tier architecture: (i) a Perception Flow that captures fine-grained user intent from text prompts, voice commands, and UI interactions; (ii) a Serving Flow that performs real-time agentic re-ranking and pruning of candidate items, grounded in a persistent Semantic Profile encoding evolving user preferences; and (iii) a Self-Evolution Flow that aligns system behavior with human judgments via Direct Preference Optimization (DPO) and an LLM-as-a-Judge ensemble. Offline evaluations show that SYF’s alignment scoring module achieves 98.85\% accuracy, substantially improving over strong few-shot baselines. Large-scale online A/B experiments on production traffic further demonstrate that SYF improves feed relevance and user sentiment, indicating a practical and scalable path toward interactive, user-steerable recommendation in industrial settings.

• INDSoft Curriculum Learning for Optimizing Fresh and Generalized Recommendations
  by Arnab Bhadury, Siyan Zheng, Anlan Yu, Palaksh Rungta, Jiawei Li, Changping Meng, Dapeng Hong, Chuan He and Onkar Dalal

  Large-scale recommender systems, particularly short-form video platforms, are often bottlenecked by massive popularity feedback loops. In such environments, as models recommend popular items, they generate an overwhelming amount of skewed training data for “head” items. This creates a self-reinforcing cycle where retrieval and ranking models memorize “head” item patterns at the expense of generalizing across the vast “tail” of the catalogue. While Curriculum Learning (CL) offers a powerful mechanism to break this feedback loop by systematically exposing models to progressively more difficult and less frequent examples, its adoption in industrial recommendation has been hampered by hardware utilization inefficiencies or the needs for complicated pre-processing techniques because dynamic data rejection algorithms tend to starve hardware accelearators (TPUs/GPUs) by becoming largely CPU-bound. In this work, we introduce a scalable Soft Curriculum Learning framework designed specifically for continuous training setups within industry-scale retrieval and ranking models. By utilizing loss annealing and in-graph weight adjustments rather than rigid data filtering, we break the popularity feedback and enable dynamic curriculum pacing without sacrificing system throughput. We demonstrate empirical evidence through applications across sequence-based retrieval models (such as SASRec), two-tower retrieval models, and large-scale continuous ranking models. Online A/B tests on our short-video platform demonstrate substantial lifts in both overall user satisfaction and fresh content consumption, all without degrading model throughput.

• INDSpace Efficient Item Embeddings for Large Recommender Systems
  by Duy Nguyen, Devanshu Jain, Mike Lawrence, Steffen Rendle, Kun Su and Anushya Subbiah

  Embedding-based methods are central to modern recommender systems, but storing item embeddings at scale poses a significant memory challenge, particularly with catalogs of millions or billions of items. Common solutions, such as uniformly reducing embedding dimensions or truncating the vocabulary to popular items, can degrade recommendation quality, limit item discovery, and exacerbate popularity bias, especially harming long-tail item visibility. This paper introduces a memory-efficient hybrid embedding scheme that allocates representation capacity based on item popularity. We represent popular items with high-dimensional embeddings to capture detailed user-item interactions, while less popular items use a compressed representation. This approach allows for full vocabulary coverage within a strict memory budget. Experiments on public datasets and large-scale proprietary data demonstrate that our method significantly improves recommendation quality, particularly for long-tail items, compared to baselines with equivalent memory footprints. Our models achieve better coverage without sacrificing overall accuracy, offering a practical solution for building scalable and diverse recommender systems.

• INDStructuring and Tokenizing Distributed User Interest Context for Generative Recommendation
  by Ruizhong Qiu, Yinglong Xia, Dongqi Fu, Hanqing Zeng, Ren Chen, Xiangjun Fan, Hong Li, Hong Yan and Hanghang Tong

  Generative recommendation is an emerging paradigm that has shown promise in industrial recommendation systems, aiming to predict the next interactions of users based on their historical behavior in an autoregressive manner. At the core of generative recommendation is item tokenization to bridge item semantics and the recommendation model. However, existing methods often struggle to effectively organize and insert complex user-behavioral and item-semantic contexts simultaneously into the recommendation model: On the one hand, existing graph-based integration methods (graph serialization, graph neural networks, etc.) either suffer from scalability issues or only leverage local graph information; On the other hand, existing semantic tokenization methods typically rely on heuristics and lack a supervision signal, which may not guarantee accurate semantic representations. To address these critical limitations in user interest context modeling, we fundamentally propose G2Rec, i.e., a scalable framework that bridges holistic graph-based user co-engagement interest modeling and semantic tokenization to boost the industry-scale generative recommendation. First, we propose to construct a sparsified item-item co-engagement graph of size O(MlogM) as the item schema, where M is the total number of interactions. Second, we design a scalable ”soft” clustering algorithm with time complexity O($\rho$MlogM) per iteration to extract the distributed interest prototypes from the constructed graph, where $\rho$ is a small constant representing the sparsity of the soft cluster membership distribution (but not a hard-assigned one-to-one exclusive membership). Third, using the item interest profiles extracted from the soft clustering, we tokenize them together with the user’s interested items to train the generative sequential recommendation model. In all, our scalable framework, G2Rec, enables the recommendation model to capture holistic and semantic user interest prototypes without requiring ground-truth interests of users, providing a more comprehensive and accurate modeling of user behavior contexts in industrial sequential recommendation. Online deployment on product surfaces and extensive experiments on public datasets demonstrate the superiority of G2Rec over existing methods.

• INDTSSR-Beta: Enhancing Billion-Scale E-Commerce Semantic Retrieval via Representation-Level Interaction
  by Guohao Tan, Jiahui Wan, Tao Wen, Dongshuai Li, Xingxian Liu, Daoning Jiang, Xiaoyao Qiu, Yuliang Yan, Dan Ou, Haihong Tang and Bo Zheng

  Semantic retrieval in e-commerce search aims to identify a compact candidate set from billion-scale product catalogs with both high recall and low latency. Dual-Encoders dominate this stage due to their efficient dot-product similarity, but this formulation limits model expressiveness and fails to capture fine-grained relationships between queries and items. While prior work has explored interaction-based similarity, its additional cost often prevents deployment at industrial scale. We present TSSR-Beta (Taobao Search Semantic Retrieval Model – Beta), which improves the expressiveness of our production Dual-Encoder TSSR through a plug-in similarity module, termed the Hybrid Interaction Head. This module introduces fine-grained matching in the representation space through two complementary pathways: InteractMLP, which captures explicit matching patterns with residual MLP blocks, and InteractTrans, which models implicit cross-dimensional interactions with a Transformer layer. Their outputs are combined by a Fusion Head to produce the final similarity score. TSSR-Beta introduces only a small parameter overhead to the Dual-Encoder without changing its architecture, enabling it to be (1) pluggable, readily adapting to diverse Dual-Encoder backbones; (2) efficiently trainable, supporting large-batch contrastive learning with massive negative sampling; and (3) industrially deployable, preserving offline item pre-encoding and supporting low-latency online retrieval with Neighborhood-Aware Approximate Nearest Neighbor (NANN). Offline experiments on the Taobao Search show a +3.90pp Hitrate@500 improvement over TSSR. On public Natural Questions and WebQA, our module further improves Recall@1 by +4.60pp and +0.90pp over public Dual-Encoder, respectively. Deployed in Taobao Search, TSSR-Beta delivers low-latency billion-scale retrieval and achieves +0.63% transaction count and +2.69% GMV gains in online A/B tests.

• INDThe Text on the Creative: An Under-Exploited Ranking Modality for Short-Form Video Ads
  by Shubham Goel, Angli Liu, Antoine Simoulin, Himanshu Thakur, Qin Huang, Benjamin Au, Guy Lebanon, Sagar Chordia, Andrew Treadway, Wendy Jiang, Yiding Wen, Jianing Fu, Alan Li, Haibo Zhang, Selahattin Akkas, Calvin Ma, Zhen Zeng and Yunyu He

  Video advertisements contain rich textual signal overlaid on the creative itself – promotional copy, prices, calls-to-action, brand and product mentions — that current production ranking systems largely under-exploit. We argue this is a high-yield, low-cost ranking modality and describe a production deployment on Instagram’s ads ranking model at billion-ad daily scale. Making this practical requires three components: a specialized OCR backbone delivering near-VLM word quality at ~80× the throughput; a smart frame-extraction stage that recovers more OCR-relevant frames at half the per-video frame budget; and a feature stack, which consists of TF–IDF tokens plus 3-byte RQ-VAE semantic IDs over a domain-tuned sentence encoder, that integrates into the ranker’s existing sparse embedding tables with no dedicated dense tower. The launched bundle delivers 0.055–0.06% offline relative NE gain (≈0.12% on Reels-class surfaces) and 0.06–0.08% NE gain in a multi-week 80%-traffic online A/B.

• INDToken Factory: Efficiently Integrating Diverse Signals into Large Recommendation Models
  by Xilun Chen, Shao-Chuan Wang, Baykal Cakici, Lukasz Heldt, Lichan Hong, Raghu Keshavan, Aniruddh Nath, Li Wei and Xinyang Yi

  Large Recommendation Models (LRMs) have demonstrated promising capabilities in industry-scale recommendation tasks. However, holistically integrating traditional signals into these transformer based architectures effectively and efficiently remains a major challenge. Conventional approaches that “textualize” these signals directly or create discrete item representations often lead to excessively long prompts, substantial memory footprints, and high computational overhead. To overcome these limitations, we propose “Token Factory”, a framework designed to transform traditional signals into “soft tokens” that can be directly processed by LRMs. This approach enables efficient integration and compression of heterogeneous input features, preventing prompt length explosion while enhancing model performance. We detail the architecture of Token Factory and present experimental results validating its effectiveness in a production-scale recommendation environment.

• INDTokens are All You Need: Dual-purpose Semantic IDs for Achieving LLM-Level I/O Efficiency in Recommendation Systems
  by Baolei Li, Yiping Yuan, Yilin Zheng, Likang Yin, Ling Liu, Fabio Soldo, Romer Rosales, Xinyang Yi and Lichan Hong

  Large-scale recommendation systems face “Memory Wall” bottlenecks due to massive, dense embedding tables. While generative retrieval uses discrete tokens for IDs, high-dimensional context still relies on inefficient dense formats. Inspired by computer vision data compression, we propose Dual-purpose Semantic IDs to achieve LLM-level I/O efficiency. Our methodology uses hierarchical quantization to condense continuous embeddings into discrete Semantic IDs performing two concurrent roles: (1) Collaborative Identity: modeling user-item interactions via learnable embedding table; and (2) Content Reconstruction: using a lightweight Semantic Decoder for on-the-fly embedding approximation. This approach replaces massive vector storage with on-demand reconstruction, reducing system overhead and data footprints. We demonstrate the efficacy of our framework through offline evaluations and successful online deployment in production-scale ranking and retrieval systems at a major video sharing platform, showing that discrete tokens are indeed all you need for highly efficient, content-rich recommendation.

• INDTowards Full Candidate Interaction: A Comprehensive Comparison Network for Better Route Recommendation
  by Hanyu Guo, Chao Chen, Longfei Xu, Chengzhang Wang, Kaikui Liu and Xiangxiang Chu

  We argue that the decision-making essence of route recommendation is comparative judgment: users choose a route because it is better than alternatives in specific aspects. The decision-critical information resides in segment-level spatial differences of non-overlapping parts between routes, which is irreversibly lost through item-level feature aggregation. Existing methods, whether attention-based or pairwise ranking approaches, follow an item-first paradigm that can only infer pairwise relations indirectly from individual route representations. To address this, we propose the Comprehensive Comparison Network (CCN), which inverts the information flow by constructing explicit comparison features from non-overlapping segments between route pairs and reasoning directly in the pairwise space. CCN introduces a Comprehensive Comparison Block that enables context-aware pairwise reasoning, where the comparison between two routes is informed by how both compare against all other candidates. We further develop an interpretable Pair Scoring Network that decomposes pairwise preferences into independent physical fields, providing field-level explanations for route selection. CCN has served as the production ranking model in Amap for over two years, achieving 85.70% offline route-trajectory coverage rate and +1.2% online improvement over the previous production model. We also release a large-scale route recommendation dataset comprising 175 million users, 512 million samples, and 6 billion routes across 370 cities.

• INDTowards Generalizable and Efficient Large-Scale Generative Recommenders
  by Qiuling Xu, Ko-Jen Hsiao and Moumita Bhattacharya

  Generative recommendation models can model user behavior as sequences of events and provide a shared backbone for multiple recommendation tasks. In production, however, pre-training gains do not automatically translate into downstream application improvements: task headroom, repeated-training cost, serving latency, and item freshness all affect transfer. We describe our experience scaling a generative recommender from 2M to 1B backbone parameters, excluding embedding and decoding layers, in a production-scale title recommendation setting. Across multiple downstream tasks, we observe task-dependent scaling behavior: some tasks approach an empirical ceiling within the observed scale range, while others continue to benefit from additional capacity. This motivates using offset scaling-law fits as a diagnostic for where additional model scale may be more or less useful. We then study production constraints that arise when applying the model in practice. Frequent retraining over trillions of behavior tokens makes training and decoding efficiency important; cached serving can make the immediate next-token target stale; and newly launched titles may need to be scored from semantic metadata before collaborative ID embeddings are reliable. We address these issues with multi-token prediction for serving-latency alignment, sampled softmax and a projected decoding head for efficient repeated training, and semantic item towers with collaborative-embedding masking for cold-start adaptation. In a one-week production-shadow evaluation over 1M users, the 1B-backbone model achieves higher MRR than the 2M-backbone baseline across all reported tasks, including a 22.5\% relative gain on the lower-predictability Task A. Overall, the results support treating model scale as one component of a production transfer problem, alongside task headroom, decoding cost, serving-latency alignment, and item generalization.

• INDTowards Reliable Social A/B Testing: Spillover-Contained Clustering with Robust Post-Experiment Analysis
  by Xu Min, Zhaoxu Yang, Kaixuan Tan, Juan Yan, Xunbin Xiong, Zihao Zhu, Kaiyu Zhu, Fenglin Cui, Yang Yang, Sihua Yang and Jianhui Bu

  A/B testing is the foundation of decision-making in online platforms, yet social products often suffer from network interference: user interactions cause treatment effects to spill over into the control group. Such spillovers bias causal estimates and undermine experimental conclusions. Existing approaches face key limitations: user-level randomization ignores network structure, while cluster-based methods often rely on general-purpose clustering that is not tailored for spillover containment and has difficulty balancing unbiasedness and statistical power at scale. We propose a spillover-contained experimentation framework with two stages. In the pre-experiment stage, we build social interaction graphs and introduce a \emph{Balanced Louvain} algorithm that produces stable, size-balanced clusters while minimizing cross-cluster edges, enabling reliable cluster-based randomization. In the post-experiment stage, we develop a tailored CUPAC estimator that leverages pre-experiment behavioral covariates to reduce the variance induced by cluster-level assignment, thereby improving statistical power. Together, these components provide both structural spillover containment and robust statistical inference. We validate our approach through large-scale social sharing experiments on Kuaishou, a platform serving hundreds of millions of users. Results show that our method substantially reduces spillover and yields more accurate assessments of social strategies than traditional user-level designs, establishing a reliable and scalable framework for networked A/B testing.

• INDTransAct V2: Production System for Lifelong User Sequence Modeling at Scale
  by Xue Xia, Saurabh Joshi, Kousik Rajesh, Kangnan Li, Yangyi Lu, Nikil Pancha, Dhruvil Badani, Jiajing Xu and Pong Eksombatchai

  Modeling lifelong user action sequences in CTR prediction faces three production challenges: \textbf{computational complexity} with quadratic transformer costs as sequences extend from hundreds to tens of thousands of actions, \textbf{infrastructure overhead} from storage and network costs that scale with both sequence length and number of candidates per request, and \textbf{weak training supervision} as sequence encoders sit far from CTR prediction heads. We present TransAct V2, a production system deployed at Pinterest serving over 630 million users, that addresses these challenges through: (1) a dual-path training-serving architecture with request-level deduplication and int8 quantization achieving 1\% logging cost, (2) fused Triton kernels with pinned memory management delivering 250$\times$ p99 latency improvement, and (3) a Next Action Loss providing direct supervision for sequence modeling within CTR frameworks. Deployed in production, TransAct V2 demonstrates significant improvements in both engagement and recommendation diversity. To support industry adoption, we open-source our serving optimizations with comprehensive ablation studies. Our work provides a reproducible blueprint for practitioners building large-scale sequential recommenders.

• INDTubiFM: Unified Item, Carousel, and Search Ranking for Streaming Discovery
  by Alexandre Salle, Chenglei Niu, Suchismit Mahapatra, Xiaoxiao Chen, Suvash Sedhain, Yaqi Wang, Shervin Shahryari, Saurabh Agrawal, Qiang Chen and Michael Tamir

  Personalized discovery systems often train separate models for item ranking, carousel ranking, and search, even though these tasks expose complementary signals from the same viewer journey: watches shape carousel and item ranking, search queries reveal intent even when they do not lead to a catalog match, and watch history helps interpret search as rewatching, continuation, or new discovery. We introduce the user story, a serialized representation that turns a user’s cross-surface history—attributes, sessions, watch events with surface and carousel context, and search events—into a single token sequence. By interleaving pretrained language tokens with domain-specific event tokens, user stories let heterogeneous recommendation and search tasks be expressed as prompted next-token prediction over a shared grammar. TubiFM is one instantiation of this approach: a Llama~3.2 1B-based model trained on user stories and prompted to rank items, carousels, or search results without task-specific architectures. In offline evaluation, this single model outperforms specialist baselines across item, carousel, and search ranking. In online A/B tests, TubiFM significantly improves search total viewing time (TVT) by +3.9% and carousel TVT by +0.30%. Item ranking is statistically neutral on TVT (+0.14%), but matches a mature production stack; across all three tasks, TubiFM serves on L40S GPUs and reduces p99 ranking latency from 500ms to 200ms. These results show that shared user stories can improve discovery while simplifying ranking systems.

• INDUNIQUE: A Unified Retrieval and Ranking System for Large-Scale Feed Recommendation
  by Zhuang Liu, Yongkang Fu, Zuodong Yang, Zonggang Wu, Yuqi Lu, Shouke Qin, Shantao Li, Guangxing Chen and Maolin Wang

  Industrial mobile feed systems rely on a retrieval-ranking pipeline to serve large-scale, heterogeneous, and fast-changing content under strict latency constraints. However, existing pipelines still suffer from two critical issues: hierarchical quantization instability in candidate retrieval and information loss between separated retrieval and ranking stages. These issues hurt long-tail and cold-start recommendation and complicate efficient serving. To address them, we present UNIQUE, a unified retrieval and ranking recommendation framework with single-layer flat quantization. UNIQUE integrates generative code-based retrieval and target-aware ranking into one early-fusion architecture, enabling end-to-end training under a shared representation while preserving efficient candidate generation. A balanced quantization mechanism is further introduced to mitigate codebook imbalance and improve long-tail representation. Offline experiments evaluate UNIQUE from both retrieval and ranking perspectives, while codebook analysis shows more balanced resource allocation than hierarchical quantization. We deploy UNIQUE in the homepage feed, discovery-page, and short-video recommendation scenarios of Mobile Baidu, serving large-scale real-world traffic. Online A/B tests achieve a 0.96% gain in total watch duration and a 1.08% gain in total distribution volume, with notable improvements for new users and highly active users. Serving measurements show 89 ms P99 latency and 44.23% online inference MFU. These results show that UNIQUE provides a stable, efficient, and production-ready framework for unified retrieval and ranking in industrial recommendation.

• INDUncertainty-Aware Reward Modeling: A Large-Scale Case Study in Video Recommendation
  by Nitu Sharaff, Brian Y. C. Leung, Shawn Andrews and James Harrison

  Historically, recommendation systems have focused on maximizing precision by treating user preference as a static, predictable target. However, this approach ignores both the inherent randomness of human behavior and the model’s own varying levels of confidence. To compensate, many industrial systems rely on “reserved slots” for exploring user interests- a heuristic that typically utilizes uniform selection. This paper presents a large-scale study on the YouTube video recommendation platform, where we integrated principled exploration directly into the ranking scores of the Homepage Ranking model. To model uncertainty, we evaluate two distinct architectures: a Variational Bayesian Last Layer (VBLL) designed to capture model’s parameter uncertainty, and Quantile Regression (QR) utilized to model the variance within the target reward distribution. By strategically targeting the right exploratory candidates, both paradigms successfully break the traditional explore-exploit trade-off, by driving improvements in both user engagement and content discovery metrics. Large-scale online A/B testing reveals that both paradigms are highly viable for production, successfully balancing exploration and exploitation while driving significant improvements in overall user engagement and content discovery. By analyzing the distinct behaviors of the VBLL and QR deployments, we compare both paradigms and hypothesize how the distinct uncertainties being modeled affect the final reward distributions. Finally, we discuss our ongoing efforts to merge these two paradigms and share preliminary notes on their integration.

• INDUniPinRec: Unifying Generative Retrieval and Ranking at Pinterest Scale
  by Hanyu Li, Yi-Ping Hsu, Aditya Mantha, Prabhat Agarwal, Laksh Bhasin, Jialu Wang, Hongtao Lin, Bella Huang, Yaxin Li, Xinyi Li, Chuxi Wang, Kousik Rajesh, Hooshmand Shokri Razaghi, Shunyao Li, Zongyue Qin, Jaewon Yang, James Li, Dhruvil Deven Badani, Jiajing Xu and Charles Rosenberg

  Modern recommendation systems predominantly train retrieval and ranking as separate models despite both increasingly relying on large transformers encoding the same user behavior data, duplicating parameters, compute, and serving cost. Prior work unifies the model architecture but not the full pipeline: input formats, training procedures, and serving stacks remain fragmented across stages. We present UniPinRec, which achieves full-stack unification of retrieval and ranking at Pinterest: one input format, one model, one training stage, deployed within existing serving infrastructure. A shared transformer encodes the user action sequence into candidate-independent representations that branch into retrieval (ANN dot-product) and ranking (cross-attention) via task-specific heads. Three ideas make this work: (1) Masked Action Modeling (MAM) eliminates interleaving, enabling weight sharing without doubling context length; (2) Blended training examples pair action sequences with feedview impression slates to satisfy both objectives jointly; (3) Cross-stage KV cache sharing reuses user-history computation from retrieval for ranking, reducing total FLOPs versus serving two independent models. Deployed in the Pinterest core surfaces, UniPinRec delivers approximately +1% online engagement lift while cutting end-to-end serving latency by 11.1% and lifting QPS by 63.6%. To our knowledge, this is the first full-stack unification of retrieval and ranking, covering inputs, model, training and serving, deployed in a production recommendation system.

• INDUniShare: A Unified Framework for Joint Video and Receiver Recommendation in Social Sharing
  by Caimeng Wang, Li Chong, Dongxu Liu, Xu Min and Jianhui Bu

  Sharing behavior on short-video platforms constitutes a complex ternary interaction among the user (sharer), the video (content), and the receiver. Traditional industrial solutions often decouple this into two independent tasks: video recommendation (predicting share probability) and receiver recommendation (predicting whom to share with), leading to suboptimal performance due to isolated modeling and inadequate information utilization. To address this, we propose \textbf{UniShare}, a novel unified framework for joint sharing prediction on both video and receiver recommendation. UniShare models the share probability through an enhanced representation learning module that incorporates pre-trained GNN and multi-modal embeddings, alongside explicit bilateral interest and relationship matching. A key innovation is our joint training paradigm, which leverages signals from both tasks to mutually enhance each other, mitigating data sparsity and improving bilateral satisfaction. We also introduce \textbf{K-Share}, a large-scale real-world dataset constructed from Kuaishou platform logs to support research in this domain. Extensive offline experiments demonstrate that UniShare significantly outperforms strong baselines on both tasks. Furthermore, online A/B testing on the Kuaishou platform confirms its effectiveness, achieving significant improvements in key metrics including the number of shares (+1.95%) and receiver reply rate (+0.482%).

• INDUniTraj: Cross-Domain Long-Sequence Modeling for Commercial Recommendation
  by Xian Hu, Ming Yue, Zhixiang Feng, Junwei Pan, Junjie Zhai, Ximei Wang, Xinrui Miao, Qian Li, Xun Liu, Shangyu Zhang, Letian Wang, Hua Lu, Zijian Zeng, Chen Cai, Wei Wang, Fei Xiong, Pengfei Xiong, Jintao Zhang, Zhiyuan Wu, Chunhui Zhang, Anan Liu, Jiulong You, Chao Deng, Yuekui Yang, Shudong Huang, Dapeng Liu, Haijie Gu and Jie Jiang

  Long-sequence modeling is increasingly important in recommender systems for capturing users’ evolving and long-term interests. In advertising, however, user interaction histories are often highly sparse due to limited exposure opportunities, making ad-only behavior sequences insufficient for effective long-sequence recommendation. To address this limitation, we propose UniTraj, a practical framework that extends sequence construction beyond the advertising domain by incorporating behaviors from content-consumption scenarios, forming unified commercial trajectories across domains and scenarios. Such unified trajectories provide richer behavioral context, but also introduce substantial heterogeneity in feature taxonomy, behavioral semantics, and optimization targets. In particular, they raise three key challenges: interference among fields from different domains and scenarios, target-specific conflicts in temporal and semantic patterns, and complex high-order dependencies across heterogeneous behavioral signals. To tackle these challenges, UniTraj adopts a two-stage design. In the first stage, it combines hierarchical hard search with a decoupled embedding-based soft search module to retrieve relevant behaviors under complex feature hierarchies while reducing conflicts between retrieval and representation learning. In the second stage, it introduces several decoupled sequence modeling components, including Decoupled Side Information Temporal Interest Networks for mitigating cross-field interference, target-decoupled positional encoding and target-decoupled SASRec for capturing target-aware temporal dynamics, and Deep TIN for modeling high-order behavioral correlations. We deploy UniTraj in a large-scale online advertising system and observe consistent improvements in online business metrics across multiple commercial scenarios. The results demonstrate the effectiveness of unified cross-domain behavior modeling for long-sequence recommendation in sparse advertising environments.

• INDVersioned Late Materialization for Ultra-Long Sequence Training in Recommendation Systems at Scale
  by Guo Liang, Ge Song, Litao Deng, Jianhui Sun, Chufeng Hu, Lu Zhang, Zhen Ma, Shouwei Chen, Weiran Liu, Sarang Sreeshylan and Xiaoxuan Meng

  Modern Deep Learning Recommendation Models (DLRMs) follow scaling laws with sequence length, driving the frontier toward ultra-long User Interaction History (UIH). However, the industry-standard “Fat Row” paradigm, which pre-materializes these sequences into every training example, creates a storage and I/O wall where data infrastructure usage exceeds GPU training capacity due to data redundancy that is amplified in multi-tenant environments where models with vastly different sequence length requirements share a union dataset. We present a versioned late materialization paradigm that eliminates this redundancy by storing UIH once in a normalized, immutable tier and reconstructing sequences just-in-time during training via lightweight versioned pointers. The system ensures Online-to-Offline (O2O) consistency through a bifurcated protocol that prevents future leakage across both streaming and batch training, while a read-optimized immutable storage layer provides multi-dimensional projection pushdown for heterogeneous model tenants. Disaggregated data preprocessing with pipelined I/O prefetching and data-affinity optimizations masks the latency of training-time sequence reconstruction, keeping training throughput compute-bound by GPUs. Deployed on production DLRMs, the system reduces training data infrastructure resource usage while enabling aggressive sequence length scaling that delivers significant model quality gains, serving as the foundational data infrastructure for modern recommendation model architectures, including HSTU and ULTRA-HSTU.

• INDWHALE: A Scalable Unified Model for Recommendation with Wukong-HSTU Architecture
  by Renqin Cai, Dawei Sun, Yuanjun Yao, Zhiyong Wang, Velvin Fu, Maggie Zhuang, Yu Shi, Zhongnan Fang, Xuan Cao, Jing Qian and Rui Li

  As scalability becomes increasingly important in recommendation model, recent architectures have advanced the modeling of two broad sources of ranking signals along separate paths: non-sequence features, including user, item, context, and cross features; and sequence features from user behavior histories. Wukong and HSTU have emerged as representative scalable backbones for these paths: Wukong scales high-order non-sequence feature-interaction modeling, while HSTU scales long user-behavior sequence modeling. Despite their complementary strengths, practical architectures that combine these two types of feature modeling remain underexplored. We present WHALE, a scalable unified recommendation architecture that jointly models non-sequence and sequence features on top of Wukong and HSTU. Each WHALE layer contains a Wukong module, an HSTU module, and an attention-based fusion module in which Wukong-derived interaction representations query HSTU-derived behavior representations. This design keeps both backbones active throughout the network and enables progressive Wukong–HSTU exchange, allowing high-order feature crosses to repeatedly retrieve fine-grained evidence from long user histories. To make WHALE practical for industrial deployment, we introduce customized Triton kernels and other model-systems co-design techniques to improve training and inference efficiency. On large-scale industrial recommendation data, WHALE achieves consistent gains in offline experiments. Additionally, it delivers positive online gains with a modest serving-throughput trade-off. Overall, WHALE provides a practical example of how these two sources of information can be scalably unified in an industrial recommendation model.

• INDWhich LLM to Fine-Tune? Agent-Driven Model Selection at Scale
  by Chen Luo, Yulin Liu, Yi Liu, Xuejing Lei, Yuchen Yan, Xin Zhang, Huimin Zeng, Hongda Mao and Monica Cheng

  Open-source model hubs now host over two million public AI models, yet teams building customer-facing AI systems must still determine which model to fine-tune for production deployment—a decision that shapes the quality, latency, and cost experienced by hundreds of millions of users. At Amazon, we spent over a year iterating on this process across multiple production use cases, where model selection remained manual, slow, and heavily biased toward a small set of familiar model families despite the rapidly expanding open-source ecosystem. We show that model selection is a recommendation problem, and introduce \textsc{AgentRec}, a multi-stage retrieval-and-ranking framework that progressively narrows hundreds of candidate models using increasingly expensive but more faithful evaluation signals. Across public benchmarks and Amazon production systems, AgentRec reduces model selection from multi scientist-weeks to couple unattended GPU-hours while matching or exceeding the quality of exhaustive manual exploration. Our results suggest that, for industrial teams deploying fine-tuned LLMs at scale, model selection can evolve from an ad-hoc bottleneck into a repeatable and continuously automated system for discovering high-quality models under real-world deployment constraints.